{"name":"BenchLeader LLM benchmark catalogue","description":"240 benchmarks from 43 benchmark publishers: what each measures, status, caveat and today's leader. Rebuilt once a day (the run starts at 22:03 UTC).","as_of":"2026-10-25","licence":{"name":"CC BY-NC 4.0","url":"https://creativecommons.org/licenses/by-nc/4.0/"},"attribution":"The catalogue (descriptions, categories, status labels and caveats) is BenchLeader's, licensed CC BY-NC 4.0 (https://creativecommons.org/licenses/by-nc/4.0/): non-commercial use with attribution by link to https://www.benchleader.com. Each leader's score is the benchmark publisher's figure and remains under its publisher's terms (credited per row).","citation":"Cite the page you used, not the site: \"BenchLeader\", the page title, the data date shown on the page (\"data as of\"), and the page's canonical URL. Example: BenchLeader, “<page title>”, data as of <date>, https://www.benchleader.com/<path>.","sources":[{"name":"Artificial Analysis","url":"https://artificialanalysis.ai/leaderboards/models","attribution":"Source: Artificial Analysis (artificialanalysis.ai). Data taken from the public leaderboard."},{"name":"Epoch AI Benchmarking Hub","url":"https://epoch.ai/benchmarks","attribution":"CC BY 4.0 — Epoch AI, ‘AI Benchmarking Hub’, epoch.ai/benchmarks. Mirrored boards credit their original publishers."},{"name":"Arena (LMArena)","url":"https://arena.ai/leaderboard","attribution":"CC BY 4.0 — lmarena-ai/leaderboard-dataset on Hugging Face."},{"name":"LiveBench","url":"https://livebench.ai","attribution":"LiveBench (livebench.ai), White et al. 2024."},{"name":"Aider polyglot leaderboard","url":"https://aider.chat/docs/leaderboards/","attribution":"Apache 2.0 — Aider-AI/aider."},{"name":"SWE-bench","url":"https://www.swebench.com","attribution":"SWE-bench (Jimenez et al. 2024), swebench.com."},{"name":"ARC Prize","url":"https://arcprize.org/leaderboard","attribution":"ARC Prize Foundation, arcprize.org."},{"name":"Berkeley Function Calling Leaderboard","url":"https://gorilla.cs.berkeley.edu/leaderboard.html","attribution":"Apache 2.0 — ShishirPatil/gorilla (BFCL)."},{"name":"τ²-bench","url":"https://taubench.com","attribution":"MIT — sierra-research/tau2-bench."},{"name":"Vals AI","url":"https://www.vals.ai/benchmarks","attribution":"Vals AI (vals.ai)."},{"name":"MathArena","url":"https://matharena.ai","attribution":"MIT — eth-sri/matharena."},{"name":"Scale AI SEAL","url":"https://labs.scale.com/leaderboard","attribution":"Scale AI SEAL Leaderboards."},{"name":"Terminal-Bench","url":"https://www.tbench.ai/leaderboard","attribution":"Terminal-Bench (tbench.ai)."},{"name":"EQ-Bench","url":"https://eqbench.com","attribution":"MIT — EQ-Bench."},{"name":"MMMU","url":"https://mmmu-benchmark.github.io","attribution":"MMMU (Yue et al.)."},{"name":"Kagi LLM Benchmark","url":"https://help.kagi.com/kagi/ai/llm-benchmark.html","attribution":"Kagi (kagi.com)."},{"name":"BrowseComp-Plus","url":"https://huggingface.co/spaces/Tevatron/BrowseComp-Plus","attribution":"BrowseComp-Plus (texttron/BrowseComp-Plus), results dataset Tevatron/BrowseComp-Plus-results."},{"name":"Kaggle Benchmarks","url":"https://www.kaggle.com/benchmarks","attribution":"Kaggle (kaggle.com); each board credits its publishing lab."},{"name":"HELM Capabilities","url":"https://crfm.stanford.edu/helm/capabilities/latest/","attribution":"Stanford Center for Research on Foundation Models (crfm.stanford.edu), Apache-2.0 results."},{"name":"models.dev","url":"https://models.dev","attribution":"MIT — sst/models.dev."},{"name":"Startrise","url":"https://www.startrise.io/benchmark/","attribution":"MIT — startriseio/sr-llm-benchmark; scores published as JSON per run."},{"name":"LiteLLM price table","url":"https://github.com/BerriAI/litellm/blob/main/model_prices_and_context_window.json","attribution":"MIT — BerriAI/litellm."},{"name":"OpenRouter","url":"https://openrouter.ai/models","attribution":"Speed and pricing data via OpenRouter (openrouter.ai)."}],"homepage":"https://www.benchleader.com/benchmarks","csv":"https://www.benchleader.com/data/benchmarks.csv","benchmarks":[{"id":"gpqa_diamond","name":"GPQA Diamond","category":"reasoning","category_label":"Reasoning","measures":"Graduate-level science questions written to be Google-proof. Run by Epoch AI.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/gpqa-diamond","models":293,"in_index":true,"index_weight":1,"status":"Saturated","caveat":"Contamination risk","leader":"Gemini 4 Argon (high) at 96.4%","leader_model_url":"https://www.benchleader.com/models/gemini-4-argon","leader_score":96.4,"unit":"percent","last_result":"2026-10-17","aliases":["gpqa"],"url":"https://www.benchleader.com/benchmarks/gpqa_diamond"},{"id":"frontiermath_t1_3","name":"FrontierMath Tiers 1–3","category":"math","category_label":"Maths","measures":"Unpublished research-level maths problems, tiers 1–3 (v2 private set). Run by Epoch AI.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/frontiermath","models":118,"in_index":true,"index_weight":1,"status":"Near-saturated","caveat":null,"leader":"GPT-6 Astra (max) at 93.7%","leader_model_url":"https://www.benchleader.com/models/gpt-6-astra","leader_score":93.68,"unit":"percent","last_result":"2026-10-17","aliases":["frontier math","frontiermath"],"url":"https://www.benchleader.com/benchmarks/frontiermath_t1_3"},{"id":"frontiermath_t4","name":"FrontierMath Tier 4","category":"math","category_label":"Maths","measures":"The hardest FrontierMath tier: problems that take expert mathematicians days. Run by Epoch AI.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/frontiermath","models":74,"in_index":true,"index_weight":0.5,"status":"Saturated","caveat":null,"leader":"GPT-6.1 Sol (max) at 100.0%","leader_model_url":"https://www.benchleader.com/models/gpt-6-1-sol","leader_score":100,"unit":"percent","last_result":"2026-10-17","aliases":["frontier math","frontiermath"],"url":"https://www.benchleader.com/benchmarks/frontiermath_t4"},{"id":"otis_mock_aime","name":"OTIS Mock AIME","category":"math","category_label":"Maths","measures":"Mock AIME competition problems from 2024–2025. Run by Epoch AI.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/otis-mock-aime-2024-2025","models":276,"in_index":true,"index_weight":0.5,"status":"Saturated","caveat":"Contamination risk","leader":"Claude Opus 5.5 (max) at 100.0%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":100,"unit":"percent","last_result":"2026-10-17","aliases":["aime"],"url":"https://www.benchleader.com/benchmarks/otis_mock_aime"},{"id":"math_level_5","name":"MATH Level 5","category":"math","category_label":"Maths","measures":"Hardest tier of the MATH dataset. Largely saturated, kept at half weight. Run by Epoch AI.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/math-level-5","models":96,"in_index":true,"index_weight":0.5,"status":"Retired","caveat":"Contamination risk","leader":"GPT-5 (high) at 98.1%","leader_model_url":"https://www.benchleader.com/models/gpt-5","leader_score":98.13,"unit":"percent","last_result":"2025-10-15","aliases":[],"url":"https://www.benchleader.com/benchmarks/math_level_5"},{"id":"swe_bench_verified_epoch","name":"SWE-bench Verified (Epoch)","category":"coding","category_label":"Coding","measures":"Epoch AI's own SWE-bench Verified runs with a constant scaffold.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/swe-bench-verified","models":33,"in_index":true,"index_weight":0.5,"status":"Active","caveat":"Harness-dependent","leader":"Claude Opus 4.7 (max) at 83.5%","leader_model_url":"https://www.benchleader.com/models/claude-opus-4-7","leader_score":83.47,"unit":"percent","last_result":"2026-06-16","aliases":[],"url":"https://www.benchleader.com/benchmarks/swe_bench_verified_epoch"},{"id":"simpleqa_verified","name":"SimpleQA Verified","category":"knowledge","category_label":"Knowledge","measures":"Short factual questions; measures recall without hallucination. Run by Epoch AI.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/simpleqa-verified","models":87,"in_index":true,"index_weight":1,"status":"Active","caveat":null,"leader":"Gemini 4 Argon (high) at 76.0%","leader_model_url":"https://www.benchleader.com/models/gemini-4-argon","leader_score":76,"unit":"percent","last_result":"2026-10-17","aliases":["simpleqa"],"url":"https://www.benchleader.com/benchmarks/simpleqa_verified"},{"id":"hle","name":"Humanity's Last Exam","category":"reasoning","category_label":"Reasoning","measures":"2,500 expert-written questions across 100+ subjects (Scale AI / CAIS).","publisher":"Scale AI / CAIS","publisher_url":"https://scale.com/leaderboard/humanitys_last_exam","models":48,"in_index":true,"index_weight":1,"status":"Active","caveat":null,"leader":"GPT-6 Astra at 54.8%","leader_model_url":"https://www.benchleader.com/models/gpt-6-astra","leader_score":54.8,"unit":"percent","last_result":"2026-10-07","aliases":["humanity's last exam","humanitys last exam"],"url":"https://www.benchleader.com/benchmarks/hle"},{"id":"terminal_bench","name":"Terminal-Bench","category":"agentic","category_label":"Agents & tools","measures":"Command-line agent tasks; best result per model across harnesses.","publisher":"Terminal-Bench","publisher_url":"https://www.tbench.ai/leaderboard","models":84,"in_index":true,"index_weight":1,"status":"Active","caveat":"Harness-dependent","leader":"GPT-5.5 at 84.7%","leader_model_url":"https://www.benchleader.com/models/gpt-5-5","leader_score":84.72,"unit":"percent","last_result":"2026-09-29","aliases":["terminalbench","tbench"],"url":"https://www.benchleader.com/benchmarks/terminal_bench"},{"id":"simplebench","name":"SimpleBench","category":"reasoning","category_label":"Reasoning","measures":"Trick questions where humans score ~84%: spatio-temporal reasoning and social intelligence.","publisher":"SimpleBench","publisher_url":"https://simple-bench.com","models":87,"in_index":true,"index_weight":1,"status":"Active","caveat":null,"leader":"Claude Fable 5 (max) at 81.9%","leader_model_url":"https://www.benchleader.com/models/claude-fable-5","leader_score":81.9,"unit":"percent","last_result":"2026-08-12","aliases":[],"url":"https://www.benchleader.com/benchmarks/simplebench"},{"id":"osworld_2","name":"OSWorld-Verified 2.0","category":"agentic","category_label":"Agents & tools","measures":"Computer-use tasks in a real desktop environment.","publisher":"OSWorld","publisher_url":"https://osworld-v1.xlang.ai/","models":14,"in_index":true,"index_weight":0.5,"status":"Active","caveat":"Harness-dependent","leader":"Claude Opus 5 (max) at 31.4%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5","leader_score":31.43,"unit":"percent","last_result":"2026-07-24","aliases":[],"url":"https://www.benchleader.com/benchmarks/osworld_2"},{"id":"gdpval","name":"GDPval","category":"agentic","category_label":"Agents & tools","measures":"Win rate against industry professionals on economically valuable tasks (OpenAI).","publisher":"OpenAI","publisher_url":"https://openai.com/index/gdpval/","models":11,"in_index":true,"index_weight":0.5,"status":"Active","caveat":"Judge-based","leader":"GPT-5.2 (no reasoning) at 49.7%","leader_model_url":"https://www.benchleader.com/models/gpt-5-2","leader_score":49.7,"unit":"percent","last_result":"2025-12-11","aliases":[],"url":"https://www.benchleader.com/benchmarks/gdpval"},{"id":"fiction_livebench","name":"Fiction.LiveBench 120k","category":"long_context","category_label":"Long context","measures":"Long-context comprehension of fiction at 120k tokens.","publisher":"Fiction.live","publisher_url":"https://fiction.live/stories/Fiction-liveBench-Mar-14-2025/oQdzQvKHw8JyXbN87","models":35,"in_index":true,"index_weight":1,"status":"Near-saturated","caveat":null,"leader":"o3 (medium) at 100.0%","leader_model_url":"https://www.benchleader.com/models/o3","leader_score":100,"unit":"percent","last_result":"2026-01-27","aliases":[],"url":"https://www.benchleader.com/benchmarks/fiction_livebench"},{"id":"scicode","name":"SciCode","category":"coding","category_label":"Coding","measures":"Research-grade scientific coding problems.","publisher":"SciCode","publisher_url":"https://scicode-bench.github.io","models":216,"in_index":true,"index_weight":0.5,"status":"Active","caveat":null,"leader":"Claude Opus 5.5 (max) at 66.9%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":66.9,"unit":"percent","last_result":"2026-10-15","aliases":[],"url":"https://www.benchleader.com/benchmarks/scicode"},{"id":"cybench","name":"Cybench","category":"agentic","category_label":"Agents & tools","measures":"Unguided capture-the-flag cybersecurity tasks.","publisher":"Cybench","publisher_url":"https://cybench.github.io","models":21,"in_index":true,"index_weight":0.5,"status":"Near-saturated","caveat":"Harness-dependent","leader":"Claude Opus 4.6 at 93.0%","leader_model_url":"https://www.benchleader.com/models/claude-opus-4-6","leader_score":93,"unit":"percent","last_result":"2026-02-05","aliases":[],"url":"https://www.benchleader.com/benchmarks/cybench"},{"id":"rli","name":"Remote Labor Index","category":"agentic","category_label":"Agents & tools","measures":"Real freelance projects completed end to end (Scale AI / CAIS).","publisher":"Scale AI / CAIS","publisher_url":"https://www.remotelabor.ai","models":16,"in_index":true,"index_weight":0.5,"status":"Active","caveat":"Harness-dependent","leader":"Claude Opus 5.5 at 21.3%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":21.25,"unit":"percent","last_result":"2026-09-22","aliases":[],"url":"https://www.benchleader.com/benchmarks/rli"},{"id":"weirdml","name":"WeirdML","category":"coding","category_label":"Coding","measures":"Unusual machine-learning tasks that require writing working training code.","publisher":"WeirdML","publisher_url":"https://htihle.github.io/weirdml.html","models":156,"in_index":true,"index_weight":0.5,"status":"Saturated","caveat":null,"leader":"GPT-6 Astra Pro (max) at 93.6%","leader_model_url":"https://www.benchleader.com/models/gpt-6-astra-pro","leader_score":93.57,"unit":"percent","last_result":"2026-09-03","aliases":[],"url":"https://www.benchleader.com/benchmarks/weirdml"},{"id":"apex_agents","name":"APEX-Agents","category":"agentic","category_label":"Agents & tools","measures":"Long-horizon professional agent tasks (Mercor).","publisher":"Mercor","publisher_url":"https://www.mercor.com/apex","models":53,"in_index":true,"index_weight":0.5,"status":"Active","caveat":"Harness-dependent","leader":"Gemini 4 Argon at 82.4%","leader_model_url":"https://www.benchleader.com/models/gemini-4-argon","leader_score":82.4,"unit":"percent","last_result":"2026-10-15","aliases":[],"url":"https://www.benchleader.com/benchmarks/apex_agents"},{"id":"frontiercode","name":"FrontierCode","category":"coding","category_label":"Coding","measures":"Hard real-world coding tasks run in agent harnesses (Cognition).","publisher":"Cognition","publisher_url":"https://cognition.com/frontiercode","models":45,"in_index":true,"index_weight":0.5,"status":"Active","caveat":null,"leader":"Claude Opus 5.5 (medium) at 54.6%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":54.64,"unit":"percent","last_result":"2026-10-07","aliases":[],"url":"https://www.benchleader.com/benchmarks/frontiercode"},{"id":"proofbench","name":"ProofBench","category":"math","category_label":"Maths","measures":"Full written proofs graded for rigour (Vals AI).","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/proof_bench","models":78,"in_index":true,"index_weight":0.5,"status":"Saturated","caveat":"Judge-based","leader":"Claude Opus 5.5 (max) at 100.0%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":100,"unit":"percent","last_result":"2026-10-12","aliases":[],"url":"https://www.benchleader.com/benchmarks/proofbench"},{"id":"gso_bench","name":"GSO-Bench","category":"coding","category_label":"Coding","measures":"Software optimisation tasks measured against expert speed-ups.","publisher":"GSO-Bench","publisher_url":"https://gso-bench.github.io","models":35,"in_index":true,"index_weight":0.5,"status":"Active","caveat":null,"leader":"Claude Fable 5.1 at 88.2%","leader_model_url":"https://www.benchleader.com/models/claude-fable-5-1","leader_score":88.24,"unit":"percent","last_result":"2026-09-03","aliases":[],"url":"https://www.benchleader.com/benchmarks/gso_bench"},{"id":"epoch_eci","name":"Epoch Capabilities Index","category":"composite","category_label":"Composite","measures":"Epoch AI's own composite capability score fitted across many benchmarks. Shown for reference; not part of the BenchLeader Index.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/eci","models":245,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Claude Opus 5.5 at 168.8","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":168.83,"unit":"score","last_result":"2026-10-15","aliases":[],"url":"https://www.benchleader.com/benchmarks/epoch_eci"},{"id":"lmarena_text","name":"Arena Text","category":"human_preference","category_label":"Human preference","measures":"Overall text arena rating with style control.","publisher":"Arena (LMArena)","publisher_url":"https://arena.ai/leaderboard/chat/text","models":373,"in_index":true,"index_weight":1,"status":"Active","caveat":"Judge-based","leader":"Gemini 4 Argon (high) at 1517","leader_model_url":"https://www.benchleader.com/models/gemini-4-argon","leader_score":1517.2,"unit":"elo","last_result":"2026-10-22","aliases":["arena","chatbot arena","lmsys"],"url":"https://www.benchleader.com/benchmarks/lmarena_text"},{"id":"lmarena_text_hard","name":"Arena Hard Prompts","category":"reasoning","category_label":"Reasoning","measures":"Text arena rating on prompts judged hard.","publisher":"Arena (LMArena)","publisher_url":"https://arena.ai/leaderboard/chat/text/hard-prompts","models":373,"in_index":true,"index_weight":0.5,"status":"Active","caveat":"Judge-based","leader":"Gemini 4 Argon (high) at 1540","leader_model_url":"https://www.benchleader.com/models/gemini-4-argon","leader_score":1540.3,"unit":"elo","last_result":"2026-10-22","aliases":["arena","hard prompts"],"url":"https://www.benchleader.com/benchmarks/lmarena_text_hard"},{"id":"lmarena_text_coding","name":"Arena Coding","category":"coding","category_label":"Coding","measures":"Text arena rating on coding prompts.","publisher":"Arena (LMArena)","publisher_url":"https://arena.ai/leaderboard/chat/text/coding","models":368,"in_index":true,"index_weight":0.5,"status":"Active","caveat":"Judge-based","leader":"Gemini 4 Argon (high) at 1552","leader_model_url":"https://www.benchleader.com/models/gemini-4-argon","leader_score":1551.6,"unit":"elo","last_result":"2026-10-22","aliases":["arena"],"url":"https://www.benchleader.com/benchmarks/lmarena_text_coding"},{"id":"lmarena_webdev","name":"Arena WebDev","category":"coding","category_label":"Coding","measures":"Pairwise votes on generated web apps.","publisher":"Arena (LMArena)","publisher_url":"https://arena.ai/leaderboard/code/webdev","models":139,"in_index":true,"index_weight":1,"status":"Active","caveat":"Judge-based","leader":"Claude Opus 5.5 (max) at 1807","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":1806.9,"unit":"elo","last_result":"2026-10-23","aliases":["arena","webdev"],"url":"https://www.benchleader.com/benchmarks/lmarena_webdev"},{"id":"lmarena_vision","name":"Arena Vision","category":"multimodal","category_label":"Multimodal","measures":"Pairwise votes on image-understanding prompts.","publisher":"Arena (LMArena)","publisher_url":"https://arena.ai/leaderboard/chat/vision","models":159,"in_index":true,"index_weight":1,"status":"Active","caveat":"Judge-based","leader":"Claude Fable 5 (high) at 1325","leader_model_url":"https://www.benchleader.com/models/claude-fable-5","leader_score":1324.9,"unit":"elo","last_result":"2026-10-19","aliases":["arena"],"url":"https://www.benchleader.com/benchmarks/lmarena_vision"},{"id":"lmarena_agent","name":"Arena Agent","category":"agentic","category_label":"Agents & tools","measures":"Task-outcome score (0–100) from real agentic sessions judged by their users.","publisher":"Arena (LMArena)","publisher_url":"https://arena.ai/leaderboard/agent","models":47,"in_index":true,"index_weight":0.5,"status":"Active","caveat":"Judge-based","leader":"Claude Opus 5.5 (max) at 20.8","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":20.8,"unit":"score","last_result":"2026-10-23","aliases":[],"url":"https://www.benchleader.com/benchmarks/lmarena_agent"},{"id":"livebench2_global","name":"LiveBench 2.0","category":"composite","category_label":"Composite","measures":"Average of LiveBench 2.0's seven suite scores (long agentic, reasoning, coding, browser use, data analysis, research, design), on the current release.","publisher":"LiveBench","publisher_url":"https://livebench.ai","models":52,"in_index":true,"index_weight":1,"status":"Near-saturated","caveat":null,"leader":"Claude Fable 5.1 at 88.0%","leader_model_url":"https://www.benchleader.com/models/claude-fable-5-1","leader_score":87.99,"unit":"percent","last_result":"2026-10-14","aliases":[],"url":"https://www.benchleader.com/benchmarks/livebench2_global"},{"id":"livebench2_reasoning","name":"LiveBench 2.0 Reasoning","category":"reasoning","category_label":"Reasoning","measures":"LiveBench 2.0 reasoning suite: short logic, spatial and theory-of-mind questions with one correct answer.","publisher":"LiveBench","publisher_url":"https://livebench.ai","models":52,"in_index":false,"index_weight":null,"status":"Saturated","caveat":null,"leader":"Gemini 4 Argon at 95.1%","leader_model_url":"https://www.benchleader.com/models/gemini-4-argon","leader_score":95.1,"unit":"percent","last_result":"2026-10-14","aliases":[],"url":"https://www.benchleader.com/benchmarks/livebench2_reasoning"},{"id":"livebench2_coding","name":"LiveBench 2.0 Coding","category":"coding","category_label":"Coding","measures":"LiveBench 2.0 coding suite: real merged-PR bug fixes on a frozen monorepo, graded deterministically.","publisher":"LiveBench","publisher_url":"https://livebench.ai","models":52,"in_index":false,"index_weight":null,"status":"Saturated","caveat":null,"leader":"Claude Fable 5 at 98.3%","leader_model_url":"https://www.benchleader.com/models/claude-fable-5","leader_score":98.3,"unit":"percent","last_result":"2026-10-14","aliases":[],"url":"https://www.benchleader.com/benchmarks/livebench2_coding"},{"id":"livebench_long_agentic","name":"LiveBench Long Agentic","category":"agentic","category_label":"Agents & tools","measures":"LiveBench 2.0 long-horizon agentic suite.","publisher":"LiveBench","publisher_url":"https://livebench.ai","models":52,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"Claude Opus 5.5 at 85.0%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":85,"unit":"percent","last_result":"2026-10-14","aliases":[],"url":"https://www.benchleader.com/benchmarks/livebench_long_agentic"},{"id":"livebench_browser","name":"LiveBench Browser","category":"agentic","category_label":"Agents & tools","measures":"LiveBench 2.0 browser-use suite.","publisher":"LiveBench","publisher_url":"https://livebench.ai","models":52,"in_index":false,"index_weight":null,"status":"Saturated","caveat":"Harness-dependent","leader":"Claude Fable 5.1 at 100.0%","leader_model_url":"https://www.benchleader.com/models/claude-fable-5-1","leader_score":100,"unit":"percent","last_result":"2026-10-14","aliases":[],"url":"https://www.benchleader.com/benchmarks/livebench_browser"},{"id":"livebench2_data","name":"LiveBench 2.0 Data Analysis","category":"knowledge","category_label":"Knowledge","measures":"LiveBench 2.0 data analysis suite: short questions on small tables with one correct answer.","publisher":"LiveBench","publisher_url":"https://livebench.ai","models":52,"in_index":false,"index_weight":null,"status":"Near-saturated","caveat":null,"leader":"Gemini 4 Argon at 85.2%","leader_model_url":"https://www.benchleader.com/models/gemini-4-argon","leader_score":85.2,"unit":"percent","last_result":"2026-10-14","aliases":[],"url":"https://www.benchleader.com/benchmarks/livebench2_data"},{"id":"livebench_research","name":"LiveBench Research","category":"knowledge","category_label":"Knowledge","measures":"LiveBench 2.0 research suite.","publisher":"LiveBench","publisher_url":"https://livebench.ai","models":52,"in_index":false,"index_weight":null,"status":"Near-saturated","caveat":null,"leader":"Claude Opus 5 at 88.9%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5","leader_score":88.9,"unit":"percent","last_result":"2026-10-14","aliases":[],"url":"https://www.benchleader.com/benchmarks/livebench_research"},{"id":"livebench_design","name":"LiveBench Design","category":"multimodal","category_label":"Multimodal","measures":"LiveBench 2.0 design suite.","publisher":"LiveBench","publisher_url":"https://livebench.ai","models":52,"in_index":false,"index_weight":null,"status":"Near-saturated","caveat":null,"leader":"Claude Opus 5.5 at 86.9%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":86.9,"unit":"percent","last_result":"2026-10-14","aliases":[],"url":"https://www.benchleader.com/benchmarks/livebench_design"},{"id":"livebench_global","name":"LiveBench 1.0","category":"composite","category_label":"Composite","measures":"Average of LiveBench 1.0's seven category scores, frozen at its final release (25 June 2026). LiveBench 2.0 replaced it.","publisher":"LiveBench","publisher_url":"https://livebench.ai/#/release/2026-06-25","models":67,"in_index":false,"index_weight":null,"status":"Near-saturated","caveat":null,"leader":"Claude Fable 5.1 (max) at 83.4%","leader_model_url":"https://www.benchleader.com/models/claude-fable-5-1","leader_score":83.41,"unit":"percent","last_result":"2026-06-25","aliases":[],"url":"https://www.benchleader.com/benchmarks/livebench_global"},{"id":"livebench_reasoning","name":"LiveBench 1.0 Reasoning","category":"reasoning","category_label":"Reasoning","measures":"Reasoning category of LiveBench 1.0, frozen at its final release (25 June 2026). LiveBench 2.0's Reasoning suite is a different test.","publisher":"LiveBench","publisher_url":"https://livebench.ai/#/release/2026-06-25","models":67,"in_index":false,"index_weight":null,"status":"Saturated","caveat":null,"leader":"GPT-6 Astra (max) at 92.7%","leader_model_url":"https://www.benchleader.com/models/gpt-6-astra","leader_score":92.65,"unit":"percent","last_result":"2026-06-25","aliases":[],"url":"https://www.benchleader.com/benchmarks/livebench_reasoning"},{"id":"livebench_coding","name":"LiveBench 1.0 Coding","category":"coding","category_label":"Coding","measures":"Coding category of LiveBench 1.0, frozen at its final release (25 June 2026). LiveBench 2.0's Coding suite is a different test.","publisher":"LiveBench","publisher_url":"https://livebench.ai/#/release/2026-06-25","models":67,"in_index":false,"index_weight":null,"status":"Near-saturated","caveat":null,"leader":"Claude Sonnet 5.5 (max) at 91.4%","leader_model_url":"https://www.benchleader.com/models/claude-sonnet-5-5","leader_score":91.37,"unit":"percent","last_result":"2026-06-25","aliases":[],"url":"https://www.benchleader.com/benchmarks/livebench_coding"},{"id":"livebench_data","name":"LiveBench 1.0 Data Analysis","category":"knowledge","category_label":"Knowledge","measures":"Data analysis category of LiveBench 1.0, frozen at its final release (25 June 2026). LiveBench 2.0's Data Analysis suite is a different test.","publisher":"LiveBench","publisher_url":"https://livebench.ai/#/release/2026-06-25","models":67,"in_index":false,"index_weight":null,"status":"Near-saturated","caveat":null,"leader":"GPT-6 Astra (max) at 83.0%","leader_model_url":"https://www.benchleader.com/models/gpt-6-astra","leader_score":82.97,"unit":"percent","last_result":"2026-06-25","aliases":[],"url":"https://www.benchleader.com/benchmarks/livebench_data"},{"id":"livebench_agentic_coding","name":"LiveBench Agentic Coding","category":"agentic","category_label":"Agents & tools","measures":"Agentic coding category of LiveBench 1.0, frozen at its final release (25 June 2026). LiveBench 2.0 dropped it.","publisher":"LiveBench","publisher_url":"https://livebench.ai/#/release/2026-06-25","models":67,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"DeepSeek V4.1 Flash (max) at 77.3%","leader_model_url":"https://www.benchleader.com/models/deepseek-v4-1-flash","leader_score":77.27,"unit":"percent","last_result":"2026-06-25","aliases":[],"url":"https://www.benchleader.com/benchmarks/livebench_agentic_coding"},{"id":"livebench_math","name":"LiveBench Mathematics","category":"math","category_label":"Maths","measures":"Mathematics category of LiveBench 1.0, frozen at its final release (25 June 2026). LiveBench 2.0 dropped it.","publisher":"LiveBench","publisher_url":"https://livebench.ai/#/release/2026-06-25","models":67,"in_index":false,"index_weight":null,"status":"Saturated","caveat":null,"leader":"Claude Opus 5.5 (max) at 97.1%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":97.08,"unit":"percent","last_result":"2026-06-25","aliases":[],"url":"https://www.benchleader.com/benchmarks/livebench_math"},{"id":"livebench_language","name":"LiveBench Language","category":"instruction","category_label":"Instruction following","measures":"Language category of LiveBench 1.0, frozen at its final release (25 June 2026). LiveBench 2.0 dropped it.","publisher":"LiveBench","publisher_url":"https://livebench.ai/#/release/2026-06-25","models":67,"in_index":false,"index_weight":null,"status":"Near-saturated","caveat":null,"leader":"Claude Fable 5 (max) at 90.7%","leader_model_url":"https://www.benchleader.com/models/claude-fable-5","leader_score":90.68,"unit":"percent","last_result":"2026-06-25","aliases":[],"url":"https://www.benchleader.com/benchmarks/livebench_language"},{"id":"livebench_if","name":"LiveBench Instruction Following","category":"instruction","category_label":"Instruction following","measures":"Instruction-following category of LiveBench 1.0, frozen at its final release (25 June 2026). LiveBench 2.0 dropped it.","publisher":"LiveBench","publisher_url":"https://livebench.ai/#/release/2026-06-25","models":67,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Gemini 3.8 Flash (high) at 81.4%","leader_model_url":"https://www.benchleader.com/models/gemini-3-8-flash","leader_score":81.41,"unit":"percent","last_result":"2026-06-25","aliases":[],"url":"https://www.benchleader.com/benchmarks/livebench_if"},{"id":"aa_intelligence_index","name":"AA Intelligence Index v4.3.2","category":"composite","category_label":"Composite","measures":"Artificial Analysis' composite of its own evaluations (GPQA, HLE, IFBench, LCR, τ²-Bench, Terminal-Bench Hard, SciCode, AA-Omniscience and more).","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/leaderboards/models","models":633,"in_index":true,"index_weight":1,"status":"Active","caveat":null,"leader":"Claude Opus 5.5 (max) at 57.6","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":57.62,"unit":"index","last_result":"2026-10-23","aliases":["artificial analysis","aa index"],"url":"https://www.benchleader.com/benchmarks/aa_intelligence_index"},{"id":"aa_ifbench","name":"IFBench","category":"instruction","category_label":"Instruction following","measures":"Precise instruction following on unseen constraints. Run by Artificial Analysis.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/ifbench","models":423,"in_index":true,"index_weight":1,"status":"Near-saturated","caveat":null,"leader":"Grok 4.3 (medium) at 83.3%","leader_model_url":"https://www.benchleader.com/models/grok-4-3","leader_score":83.33,"unit":"percent","last_result":"2026-10-23","aliases":[],"url":"https://www.benchleader.com/benchmarks/aa_ifbench"},{"id":"aa_lcr","name":"AA-LCR","category":"long_context","category_label":"Long context","measures":"Long-context reasoning across ~100k-token document sets. Run by Artificial Analysis.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/artificial-analysis-long-context-reasoning","models":548,"in_index":true,"index_weight":1,"status":"Near-saturated","caveat":null,"leader":"Kimi K3 (max) at 88.7%","leader_model_url":"https://www.benchleader.com/models/kimi-k3","leader_score":88.67,"unit":"percent","last_result":"2026-10-23","aliases":[],"url":"https://www.benchleader.com/benchmarks/aa_lcr"},{"id":"aa_mmmu_pro","name":"MMMU-Pro","category":"multimodal","category_label":"Multimodal","measures":"College-level multimodal understanding. Run by Artificial Analysis.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/mmmu-pro","models":282,"in_index":true,"index_weight":1,"status":"Near-saturated","caveat":null,"leader":"Claude Opus 5.5 (max) at 87.7%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":87.69,"unit":"percent","last_result":"2026-10-23","aliases":[],"url":"https://www.benchleader.com/benchmarks/aa_mmmu_pro"},{"id":"aa_omniscience","name":"AA-Omniscience","category":"knowledge","category_label":"Knowledge","measures":"Knowledge recall that penalises hallucination: correct answers minus confident wrong ones, from −100 to 100. Run by Artificial Analysis.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/omniscience","models":552,"in_index":true,"index_weight":1,"status":"Active","caveat":null,"leader":"Claude Opus 5.5 (max) at 46.4","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":46.42,"unit":"score","last_result":"2026-10-23","aliases":[],"url":"https://www.benchleader.com/benchmarks/aa_omniscience"},{"id":"aa_terminalbench_hard","name":"Terminal-Bench Hard","category":"agentic","category_label":"Agents & tools","measures":"Hard subset of Terminal-Bench in a fixed harness. Run by Artificial Analysis.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/terminalbench-hard","models":406,"in_index":true,"index_weight":0.5,"status":"Active","caveat":"Harness-dependent","leader":"GPT-5.6 Sol (max) at 65.9%","leader_model_url":"https://www.benchleader.com/models/gpt-5-6-sol","leader_score":65.91,"unit":"percent","last_result":"2026-07-09","aliases":["terminalbench","tbench"],"url":"https://www.benchleader.com/benchmarks/aa_terminalbench_hard"},{"id":"aa_gpqa","name":"GPQA Diamond (AA)","category":"reasoning","category_label":"Reasoning","measures":"Artificial Analysis' own GPQA Diamond run. Shown for reference; Epoch's run is in the index.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/gpqa-diamond","models":559,"in_index":false,"index_weight":null,"status":"Saturated","caveat":"Contamination risk","leader":"GPT-6 Astra (xhigh) at 96.3%","leader_model_url":"https://www.benchleader.com/models/gpt-6-astra","leader_score":96.26,"unit":"percent","last_result":"2026-10-15","aliases":[],"url":"https://www.benchleader.com/benchmarks/aa_gpqa"},{"id":"aa_hle","name":"Humanity's Last Exam (AA)","category":"reasoning","category_label":"Reasoning","measures":"Artificial Analysis' own HLE run. Shown for reference.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/humanitys-last-exam","models":615,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Claude Opus 5.5 (max) at 61.4%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":61.35,"unit":"percent","last_result":"2026-10-23","aliases":[],"url":"https://www.benchleader.com/benchmarks/aa_hle"},{"id":"aa_scicode","name":"SciCode (AA)","category":"coding","category_label":"Coding","measures":"Artificial Analysis' own SciCode run. Shown for reference.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/scicode","models":232,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Claude Opus 5.5 (max) at 66.9%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":66.9,"unit":"percent","last_result":"2026-10-23","aliases":[],"url":"https://www.benchleader.com/benchmarks/aa_scicode"},{"id":"aa_tau2","name":"τ²-Bench Telecom (AA)","category":"agentic","category_label":"Agents & tools","measures":"Artificial Analysis' τ²-Bench telecom run. Shown for reference.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/tau2-bench","models":413,"in_index":false,"index_weight":null,"status":"Saturated","caveat":"Harness-dependent","leader":"GLM 5.2 (max) at 99.1%","leader_model_url":"https://www.benchleader.com/models/glm-5-2","leader_score":99.12,"unit":"percent","last_result":"2026-07-09","aliases":["tau","tau2","taubench","tau bench","tau-bench","tau2 bench","tau2-bench"],"url":"https://www.benchleader.com/benchmarks/aa_tau2"},{"id":"vals_aime","name":"AIME (Vals)","category":"math","category_label":"Maths","measures":"Recent AIME problems, run by Vals AI at stated reasoning effort.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/aime","models":92,"in_index":true,"index_weight":0.5,"status":"Saturated","caveat":"Contamination risk","leader":"Gemini 3.1 Pro (high) at 98.1%","leader_model_url":"https://www.benchleader.com/models/gemini-3-1-pro","leader_score":98.13,"unit":"percent","last_result":"2026-04-16","aliases":["aime"],"url":"https://www.benchleader.com/benchmarks/vals_aime"},{"id":"vals_livecodebench","name":"LiveCodeBench","category":"coding","category_label":"Coding","measures":"Competitive programming problems released after model cutoffs. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/lcb","models":137,"in_index":true,"index_weight":1,"status":"Near-saturated","caveat":null,"leader":"Claude Fable 5.1 at 90.5%","leader_model_url":"https://www.benchleader.com/models/claude-fable-5-1","leader_score":90.52,"unit":"percent","last_result":"2026-09-01","aliases":["lcb","livecode"],"url":"https://www.benchleader.com/benchmarks/vals_livecodebench"},{"id":"vals_mmlu_pro","name":"MMLU-Pro","category":"knowledge","category_label":"Knowledge","measures":"Harder, reasoning-heavy successor to MMLU. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/mmlu_pro","models":130,"in_index":true,"index_weight":1,"status":"Saturated","caveat":"Contamination risk","leader":"Claude Fable 5.1 at 92.4%","leader_model_url":"https://www.benchleader.com/models/claude-fable-5-1","leader_score":92.38,"unit":"percent","last_result":"2026-09-01","aliases":["mmlu"],"url":"https://www.benchleader.com/benchmarks/vals_mmlu_pro"},{"id":"vals_ioi","name":"IOI","category":"coding","category_label":"Coding","measures":"International Olympiad in Informatics problems. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/ioi","models":43,"in_index":true,"index_weight":0.5,"status":"Saturated","caveat":null,"leader":"Gemini 4 Argon (high) at 100.0%","leader_model_url":"https://www.benchleader.com/models/gemini-4-argon","leader_score":100,"unit":"percent","last_result":"2026-10-20","aliases":[],"url":"https://www.benchleader.com/benchmarks/vals_ioi"},{"id":"vals_legalbench","name":"LegalBench","category":"knowledge","category_label":"Knowledge","measures":"Legal reasoning tasks written by lawyers. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/legal_bench","models":143,"in_index":true,"index_weight":0.5,"status":"Near-saturated","caveat":null,"leader":"Claude Fable 5 at 88.6%","leader_model_url":"https://www.benchleader.com/models/claude-fable-5","leader_score":88.56,"unit":"percent","last_result":"2026-10-13","aliases":[],"url":"https://www.benchleader.com/benchmarks/vals_legalbench"},{"id":"vals_corpfin","name":"CorpFin","category":"knowledge","category_label":"Knowledge","measures":"Questions over real corporate finance documents. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/corp_fin_v2","models":124,"in_index":true,"index_weight":0.5,"status":"Active","caveat":null,"leader":"Claude Opus 5 at 73.2%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5","leader_score":73.19,"unit":"percent","last_result":"2026-08-12","aliases":[],"url":"https://www.benchleader.com/benchmarks/vals_corpfin"},{"id":"vals_taxeval","name":"TaxEval","category":"knowledge","category_label":"Knowledge","measures":"US tax questions checked by professionals. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/tax_eval_v2","models":138,"in_index":true,"index_weight":0.5,"status":"Active","caveat":null,"leader":"Muse Spark 1.2 (xhigh) at 80.4%","leader_model_url":"https://www.benchleader.com/models/muse-spark-1-2","leader_score":80.38,"unit":"percent","last_result":"2026-09-01","aliases":[],"url":"https://www.benchleader.com/benchmarks/vals_taxeval"},{"id":"vals_medqa","name":"MedQA","category":"knowledge","category_label":"Knowledge","measures":"US medical licensing exam questions. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/medqa","models":91,"in_index":true,"index_weight":0.5,"status":"Saturated","caveat":"Contamination risk","leader":"o1 (high) at 96.5%","leader_model_url":"https://www.benchleader.com/models/o1","leader_score":96.52,"unit":"percent","last_result":"2026-04-16","aliases":[],"url":"https://www.benchleader.com/benchmarks/vals_medqa"},{"id":"vals_mgsm","name":"MGSM","category":"math","category_label":"Maths","measures":"Grade-school maths in ten languages. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/mgsm","models":71,"in_index":true,"index_weight":0.5,"status":"Saturated","caveat":"Contamination risk","leader":"Claude Opus 4.5 (thinking) at 95.2%","leader_model_url":"https://www.benchleader.com/models/claude-opus-4-5","leader_score":95.2,"unit":"percent","last_result":"2026-01-09","aliases":[],"url":"https://www.benchleader.com/benchmarks/vals_mgsm"},{"id":"vals_terminal_bench","name":"Terminal-Bench 2.1 (Vals)","category":"agentic","category_label":"Agents & tools","measures":"Terminal-Bench 2.1 in Vals AI's fixed harness.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/terminal-bench-2-1","models":73,"in_index":true,"index_weight":0.5,"status":"Near-saturated","caveat":"Harness-dependent","leader":"Claude Opus 5.5 at 87.6%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":87.64,"unit":"percent","last_result":"2026-09-28","aliases":["terminalbench","tbench"],"url":"https://www.benchleader.com/benchmarks/vals_terminal_bench"},{"id":"vals_swebench","name":"SWE-bench (Vals)","category":"coding","category_label":"Coding","measures":"Vals AI's own SWE-bench run. Shown for reference.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/swebench","models":85,"in_index":false,"index_weight":null,"status":"Saturated","caveat":"Harness-dependent","leader":"Claude Opus 5 at 97.0%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5","leader_score":97,"unit":"percent","last_result":"2026-09-01","aliases":[],"url":"https://www.benchleader.com/benchmarks/vals_swebench"},{"id":"vals_gpqa","name":"GPQA Diamond (Vals)","category":"reasoning","category_label":"Reasoning","measures":"Vals AI's own GPQA Diamond run. Shown for reference.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/gpqa","models":130,"in_index":false,"index_weight":null,"status":"Saturated","caveat":"Contamination risk","leader":"Gemini 3.1 Pro (high) at 95.5%","leader_model_url":"https://www.benchleader.com/models/gemini-3-1-pro","leader_score":95.45,"unit":"percent","last_result":"2026-09-01","aliases":[],"url":"https://www.benchleader.com/benchmarks/vals_gpqa"},{"id":"vals_index","name":"Vals Index","category":"composite","category_label":"Composite","measures":"Vals AI's composite across its benchmarks. Shown for reference.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/vals_index","models":25,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Claude Sonnet 5.5 at 70.8","leader_model_url":"https://www.benchleader.com/models/claude-sonnet-5-5","leader_score":70.85,"unit":"index","last_result":"2026-10-24","aliases":[],"url":"https://www.benchleader.com/benchmarks/vals_index"},{"id":"matharena_aime_2026","name":"AIME 2026","category":"math","category_label":"Maths","measures":"AIME I and II 2026, evaluated on release by MathArena.","publisher":"MathArena","publisher_url":"https://matharena.ai/?comp=aime--aime_2026","models":31,"in_index":true,"index_weight":1,"status":"Saturated","caveat":null,"leader":"GPT-5.5 (xhigh) at 100.0%","leader_model_url":"https://www.benchleader.com/models/gpt-5-5","leader_score":100,"unit":"percent","last_result":null,"aliases":["aime"],"url":"https://www.benchleader.com/benchmarks/matharena_aime_2026"},{"id":"matharena_hmmt_feb_2026","name":"HMMT February 2026","category":"math","category_label":"Maths","measures":"Harvard-MIT Mathematics Tournament, February 2026. MathArena.","publisher":"MathArena","publisher_url":"https://matharena.ai/?comp=hmmt--hmmt_feb_2026","models":31,"in_index":true,"index_weight":1,"status":"Saturated","caveat":null,"leader":"GPT-5.5 (xhigh) at 98.5%","leader_model_url":"https://www.benchleader.com/models/gpt-5-5","leader_score":98.48,"unit":"percent","last_result":null,"aliases":[],"url":"https://www.benchleader.com/benchmarks/matharena_hmmt_feb_2026"},{"id":"matharena_usamo_2026","name":"USAMO 2026","category":"math","category_label":"Maths","measures":"USA Mathematical Olympiad 2026, proof-graded by MathArena. Shown for reference; too few models, too close to the ceiling.","publisher":"MathArena","publisher_url":"https://matharena.ai/?comp=usamo--usamo_2026","models":9,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"GPT-5.5 (xhigh) at 98.2%","leader_model_url":"https://www.benchleader.com/models/gpt-5-5","leader_score":98.21,"unit":"percent","last_result":null,"aliases":[],"url":"https://www.benchleader.com/benchmarks/matharena_usamo_2026"},{"id":"matharena_imo_2025","name":"IMO 2025","category":"math","category_label":"Maths","measures":"International Mathematical Olympiad 2025, proof-graded. MathArena.","publisher":"MathArena","publisher_url":"https://matharena.ai/?comp=imo--imo_2025","models":6,"in_index":true,"index_weight":0.5,"status":"Active","caveat":null,"leader":"GPT-5 (high) at 38.1%","leader_model_url":"https://www.benchleader.com/models/gpt-5","leader_score":38.1,"unit":"percent","last_result":null,"aliases":[],"url":"https://www.benchleader.com/benchmarks/matharena_imo_2025"},{"id":"matharena_apex_2025","name":"MathArena Apex","category":"math","category_label":"Maths","measures":"The hardest problems from recent competitions. MathArena.","publisher":"MathArena","publisher_url":"https://matharena.ai/?comp=apex--apex_2025","models":46,"in_index":true,"index_weight":0.5,"status":"Active","caveat":null,"leader":"Claude Opus 4.8 (max) at 81.3%","leader_model_url":"https://www.benchleader.com/models/claude-opus-4-8","leader_score":81.25,"unit":"percent","last_result":null,"aliases":[],"url":"https://www.benchleader.com/benchmarks/matharena_apex_2025"},{"id":"scale_swe_bench_pro","name":"SWE-Bench Pro","category":"coding","category_label":"Coding","measures":"Long-horizon software engineering tasks in public repositories. Scale AI.","publisher":"Scale AI SEAL","publisher_url":"https://labs.scale.com/leaderboard/swe_bench_pro","models":24,"in_index":true,"index_weight":1,"status":"Active","caveat":"Harness-dependent","leader":"Muse Spark 1.1 at 61.5%","leader_model_url":"https://www.benchleader.com/models/muse-spark-1-1","leader_score":61.5,"unit":"percent","last_result":"2026-07-09","aliases":["swebench pro","swe-bench pro"],"url":"https://www.benchleader.com/benchmarks/scale_swe_bench_pro"},{"id":"scale_mcp_atlas","name":"MCP Atlas","category":"agentic","category_label":"Agents & tools","measures":"Real-world tool use through the Model Context Protocol. Scale AI.","publisher":"Scale AI SEAL","publisher_url":"https://labs.scale.com/leaderboard/mcp_atlas","models":33,"in_index":true,"index_weight":1,"status":"Near-saturated","caveat":"Harness-dependent","leader":"Muse Spark 1.1 at 88.1%","leader_model_url":"https://www.benchleader.com/models/muse-spark-1-1","leader_score":88.1,"unit":"percent","last_result":"2026-09-17","aliases":[],"url":"https://www.benchleader.com/benchmarks/scale_mcp_atlas"},{"id":"scale_multichallenge","name":"MultiChallenge","category":"instruction","category_label":"Instruction following","measures":"Multi-turn instruction following under realistic conversation constraints. Scale AI.","publisher":"Scale AI SEAL","publisher_url":"https://labs.scale.com/leaderboard/multichallenge","models":33,"in_index":true,"index_weight":1,"status":"Active","caveat":null,"leader":"GPT-6.1 Sol at 82.0%","leader_model_url":"https://www.benchleader.com/models/gpt-6-1-sol","leader_score":81.95,"unit":"percent","last_result":"2026-10-21","aliases":[],"url":"https://www.benchleader.com/benchmarks/scale_multichallenge"},{"id":"scale_prbench_finance","name":"PRBench Finance","category":"knowledge","category_label":"Knowledge","measures":"Professional reasoning in finance, graded by practitioners. Scale AI.","publisher":"Scale AI SEAL","publisher_url":"https://labs.scale.com/leaderboard/prbench-finance","models":37,"in_index":true,"index_weight":0.5,"status":"Active","caveat":null,"leader":"Muse Spark 1.3 at 59.5%","leader_model_url":"https://www.benchleader.com/models/muse-spark-1-3","leader_score":59.54,"unit":"percent","last_result":"2026-10-16","aliases":[],"url":"https://www.benchleader.com/benchmarks/scale_prbench_finance"},{"id":"scale_prbench_legal","name":"PRBench Legal","category":"knowledge","category_label":"Knowledge","measures":"Professional reasoning in legal practice, graded by practitioners. Scale AI.","publisher":"Scale AI SEAL","publisher_url":"https://labs.scale.com/leaderboard/prbench-legal","models":42,"in_index":true,"index_weight":0.5,"status":"Active","caveat":null,"leader":"Muse Spark 1.3 at 61.6%","leader_model_url":"https://www.benchleader.com/models/muse-spark-1-3","leader_score":61.56,"unit":"percent","last_result":"2026-10-20","aliases":[],"url":"https://www.benchleader.com/benchmarks/scale_prbench_legal"},{"id":"scale_enigma_eval","name":"EnigmaEval","category":"reasoning","category_label":"Reasoning","measures":"Puzzle-hunt style multi-step reasoning. Scale AI.","publisher":"Scale AI SEAL","publisher_url":"https://labs.scale.com/leaderboard/enigma_eval","models":5,"in_index":true,"index_weight":0.5,"status":"Active","caveat":null,"leader":"Claude Fable 5 (high) at 39.3%","leader_model_url":"https://www.benchleader.com/models/claude-fable-5","leader_score":39.28,"unit":"percent","last_result":"2026-03-25","aliases":[],"url":"https://www.benchleader.com/benchmarks/scale_enigma_eval"},{"id":"scale_vista","name":"VISTA","category":"multimodal","category_label":"Multimodal","measures":"Vision-language understanding. Scale AI.","publisher":"Scale AI SEAL","publisher_url":"https://labs.scale.com/leaderboard/visual_language_understanding","models":57,"in_index":true,"index_weight":1,"status":"Active","caveat":null,"leader":"Gemini 2.5 Pro at 54.6%","leader_model_url":"https://www.benchleader.com/models/gemini-2-5-pro","leader_score":54.65,"unit":"percent","last_result":"2026-03-23","aliases":[],"url":"https://www.benchleader.com/benchmarks/scale_vista"},{"id":"scale_multinrc","name":"MultiNRC","category":"knowledge","category_label":"Knowledge","measures":"Native multilingual reasoning across languages. Scale AI.","publisher":"Scale AI SEAL","publisher_url":"https://labs.scale.com/leaderboard/multinrc","models":43,"in_index":true,"index_weight":0.5,"status":"Active","caveat":null,"leader":"Muse Spark 1.1 at 65.6%","leader_model_url":"https://www.benchleader.com/models/muse-spark-1-1","leader_score":65.59,"unit":"percent","last_result":"2026-07-09","aliases":[],"url":"https://www.benchleader.com/benchmarks/scale_multinrc"},{"id":"scale_hil","name":"HiL-Bench","category":"agentic","category_label":"Agents & tools","measures":"Whether agents notice information gaps and ask clarifying questions. Scale AI.","publisher":"Scale AI SEAL","publisher_url":"https://labs.scale.com/leaderboard/hil","models":30,"in_index":true,"index_weight":0.5,"status":"Active","caveat":"Harness-dependent","leader":"Muse Wonder at 79.7%","leader_model_url":"https://www.benchleader.com/models/muse-wonder","leader_score":79.67,"unit":"percent","last_result":"2026-10-21","aliases":[],"url":"https://www.benchleader.com/benchmarks/scale_hil"},{"id":"scale_tutorbench","name":"TutorBench","category":"instruction","category_label":"Instruction following","measures":"Tutoring tasks for high-school and AP subjects. Scale AI.","publisher":"Scale AI SEAL","publisher_url":"https://labs.scale.com/leaderboard/tutorbench","models":27,"in_index":true,"index_weight":0.5,"status":"Active","caveat":"Judge-based","leader":"Muse Spark at 68.5%","leader_model_url":"https://www.benchleader.com/models/muse-spark","leader_score":68.55,"unit":"percent","last_result":"2026-04-08","aliases":[],"url":"https://www.benchleader.com/benchmarks/scale_tutorbench"},{"id":"eqbench4","name":"EQ-Bench 4","category":"human_preference","category_label":"Human preference","measures":"Emotional intelligence in multi-turn role-play, judged pairwise into an Elo-style rating.","publisher":"EQ-Bench","publisher_url":"https://eqbench.com/","models":30,"in_index":true,"index_weight":0.5,"status":"Active","caveat":"Judge-based","leader":"Claude Opus 5 at 1360","leader_model_url":"https://www.benchleader.com/models/claude-opus-5","leader_score":1360,"unit":"elo","last_result":"2026-10-23","aliases":["eq bench","eq-bench","emotional"],"url":"https://www.benchleader.com/benchmarks/eqbench4"},{"id":"mmmu_val","name":"MMMU (validation)","category":"multimodal","category_label":"Multimodal","measures":"Official MMMU validation-set leaderboard.","publisher":"MMMU","publisher_url":"https://mmmu-benchmark.github.io/#leaderboard","models":50,"in_index":true,"index_weight":0.5,"status":"Near-saturated","caveat":"Contamination risk","leader":"GPT-5.1 at 85.4%","leader_model_url":"https://www.benchleader.com/models/gpt-5-1","leader_score":85.4,"unit":"percent","last_result":"2025-11-24","aliases":[],"url":"https://www.benchleader.com/benchmarks/mmmu_val"},{"id":"mmmu_pro_official","name":"MMMU-Pro (official)","category":"multimodal","category_label":"Multimodal","measures":"Official MMMU-Pro leaderboard. Shown for reference.","publisher":"MMMU","publisher_url":"https://mmmu-benchmark.github.io/#leaderboard","models":28,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Gemini 3 Pro at 81.0%","leader_model_url":"https://www.benchleader.com/models/gemini-3-pro","leader_score":81,"unit":"percent","last_result":"2026-04-08","aliases":[],"url":"https://www.benchleader.com/benchmarks/mmmu_pro_official"},{"id":"kagi_llm","name":"Kagi LLM Benchmark","category":"reasoning","category_label":"Reasoning","measures":"Kagi's unpublished reasoning, coding and instruction questions.","publisher":"Kagi LLM Benchmark","publisher_url":"https://help.kagi.com/kagi/ai/llm-benchmark.html","models":136,"in_index":true,"index_weight":0.5,"status":"Near-saturated","caveat":null,"leader":"Claude Fable 5 (thinking) at 91.4%","leader_model_url":"https://www.benchleader.com/models/claude-fable-5","leader_score":91.4,"unit":"percent","last_result":null,"aliases":[],"url":"https://www.benchleader.com/benchmarks/kagi_llm"},{"id":"helm_ifeval","name":"IFEval (HELM)","category":"instruction","category_label":"Instruction following","measures":"Strict instruction following on verifiable constraints. Run by Stanford CRFM's HELM.","publisher":"HELM Capabilities","publisher_url":"https://crfm.stanford.edu/helm/capabilities/latest/#/leaderboard","models":68,"in_index":true,"index_weight":1,"status":"Saturated","caveat":null,"leader":"Grok 3 mini at 95.1%","leader_model_url":"https://www.benchleader.com/models/grok-3-mini","leader_score":95.1,"unit":"percent","last_result":"2025-11-24","aliases":[],"url":"https://www.benchleader.com/benchmarks/helm_ifeval"},{"id":"helm_omni_math","name":"Omni-MATH (HELM)","category":"math","category_label":"Maths","measures":"Olympiad-level maths problems. Run by Stanford CRFM's HELM.","publisher":"HELM Capabilities","publisher_url":"https://crfm.stanford.edu/helm/capabilities/latest/#/leaderboard","models":68,"in_index":true,"index_weight":1,"status":"Active","caveat":null,"leader":"GPT-5 mini at 72.2%","leader_model_url":"https://www.benchleader.com/models/gpt-5-mini","leader_score":72.2,"unit":"percent","last_result":"2025-11-24","aliases":[],"url":"https://www.benchleader.com/benchmarks/helm_omni_math"},{"id":"helm_wildbench","name":"WildBench (HELM)","category":"instruction","category_label":"Instruction following","measures":"Real user tasks graded by a rubric-guided judge. Run by Stanford CRFM's HELM.","publisher":"HELM Capabilities","publisher_url":"https://crfm.stanford.edu/helm/capabilities/latest/#/leaderboard","models":68,"in_index":true,"index_weight":0.5,"status":"Near-saturated","caveat":"Judge-based","leader":"Qwen3 235B A22B 2507 at 86.6%","leader_model_url":"https://www.benchleader.com/models/qwen3-235b-a22b-2507","leader_score":86.6,"unit":"percent","last_result":"2025-11-24","aliases":[],"url":"https://www.benchleader.com/benchmarks/helm_wildbench"},{"id":"helm_mmlu_pro","name":"MMLU-Pro (HELM)","category":"knowledge","category_label":"Knowledge","measures":"Stanford CRFM's own MMLU-Pro run. Shown for reference; Vals AI's run is in the index.","publisher":"HELM Capabilities","publisher_url":"https://crfm.stanford.edu/helm/capabilities/latest/#/leaderboard","models":68,"in_index":false,"index_weight":null,"status":"Near-saturated","caveat":"Contamination risk","leader":"Gemini 3 Pro at 90.3%","leader_model_url":"https://www.benchleader.com/models/gemini-3-pro","leader_score":90.3,"unit":"percent","last_result":"2025-11-24","aliases":["mmlu"],"url":"https://www.benchleader.com/benchmarks/helm_mmlu_pro"},{"id":"helm_gpqa","name":"GPQA Diamond (HELM)","category":"reasoning","category_label":"Reasoning","measures":"Stanford CRFM's own GPQA Diamond run. Shown for reference; Epoch's run is in the index.","publisher":"HELM Capabilities","publisher_url":"https://crfm.stanford.edu/helm/capabilities/latest/#/leaderboard","models":68,"in_index":false,"index_weight":null,"status":"Active","caveat":"Contamination risk","leader":"Gemini 3 Pro at 80.3%","leader_model_url":"https://www.benchleader.com/models/gemini-3-pro","leader_score":80.3,"unit":"percent","last_result":"2025-11-24","aliases":[],"url":"https://www.benchleader.com/benchmarks/helm_gpqa"},{"id":"helm_mean","name":"HELM Capabilities mean","category":"composite","category_label":"Composite","measures":"Mean of HELM's five capability scenarios. Shown for reference.","publisher":"HELM Capabilities","publisher_url":"https://crfm.stanford.edu/helm/capabilities/latest/#/leaderboard","models":68,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"GPT-5 mini at 81.9%","leader_model_url":"https://www.benchleader.com/models/gpt-5-mini","leader_score":81.9,"unit":"percent","last_result":"2025-11-24","aliases":["helm","stanford"],"url":"https://www.benchleader.com/benchmarks/helm_mean"},{"id":"aider_polyglot","name":"Aider Polyglot","category":"coding","category_label":"Coding","measures":"Percent of 225 Exercism problems solved on the second attempt.","publisher":"Aider polyglot leaderboard","publisher_url":"https://aider.chat/docs/leaderboards/","models":48,"in_index":true,"index_weight":1,"status":"Retired","caveat":"Harness-dependent","leader":"GPT-5 at 88.0%","leader_model_url":"https://www.benchleader.com/models/gpt-5","leader_score":88,"unit":"percent","last_result":"2025-10-03","aliases":["aider"],"url":"https://www.benchleader.com/benchmarks/aider_polyglot"},{"id":"swe_bench_verified","name":"SWE-bench Verified (bash only)","category":"coding","category_label":"Coding","measures":"Official SWE-bench Verified results from the minimal bash-only mini-SWE-agent harness.","publisher":"SWE-bench","publisher_url":"https://www.swebench.com/#verified","models":45,"in_index":true,"index_weight":1,"status":"Active","caveat":"Harness-dependent","leader":"Claude Opus 4.5 (high) at 76.8%","leader_model_url":"https://www.benchleader.com/models/claude-opus-4-5","leader_score":76.8,"unit":"percent","last_result":"2026-02-26","aliases":["swebench","swe-bench","swe bench verified"],"url":"https://www.benchleader.com/benchmarks/swe_bench_verified"},{"id":"swe_bench_verified_any","name":"SWE-bench Verified (any scaffold)","category":"coding","category_label":"Coding","measures":"Best official SWE-bench Verified result for the model with any agent scaffold. Shown for reference.","publisher":"SWE-bench","publisher_url":"https://www.swebench.com/#verified","models":64,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"Claude Opus 4.5 at 79.2%","leader_model_url":"https://www.benchleader.com/models/claude-opus-4-5","leader_score":79.2,"unit":"percent","last_result":"2026-02-19","aliases":["swebench","swe-bench"],"url":"https://www.benchleader.com/benchmarks/swe_bench_verified_any"},{"id":"arc_agi_1","name":"ARC-AGI-1","category":"reasoning","category_label":"Reasoning","measures":"Semi-private ARC-AGI-1 evaluation set.","publisher":"ARC Prize","publisher_url":"https://arcprize.org/leaderboard","models":256,"in_index":true,"index_weight":0.5,"status":"Saturated","caveat":null,"leader":"Gemini 4 Argon (high) at 98.5%","leader_model_url":"https://www.benchleader.com/models/gemini-4-argon","leader_score":98.5,"unit":"percent","last_result":"2026-10-17","aliases":["arc","arc-agi"],"url":"https://www.benchleader.com/benchmarks/arc_agi_1"},{"id":"arc_agi_2","name":"ARC-AGI-2","category":"reasoning","category_label":"Reasoning","measures":"Semi-private ARC-AGI-2 evaluation set.","publisher":"ARC Prize","publisher_url":"https://arcprize.org/leaderboard","models":256,"in_index":true,"index_weight":1,"status":"Saturated","caveat":null,"leader":"GPT-6 Astra (max) at 95.0%","leader_model_url":"https://www.benchleader.com/models/gpt-6-astra","leader_score":95,"unit":"percent","last_result":"2026-10-17","aliases":["arc","arc-agi"],"url":"https://www.benchleader.com/benchmarks/arc_agi_2"},{"id":"arc_agi_3","name":"ARC-AGI-3","category":"reasoning","category_label":"Reasoning","measures":"Semi-private ARC-AGI-3 interactive evaluation set.","publisher":"ARC Prize","publisher_url":"https://arcprize.org/leaderboard","models":69,"in_index":true,"index_weight":0.5,"status":"Near-saturated","caveat":null,"leader":"GPT-6 Astra (high) at 99.9%","leader_model_url":"https://www.benchleader.com/models/gpt-6-astra","leader_score":99.9,"unit":"percent","last_result":"2026-10-13","aliases":["arc","arc-agi"],"url":"https://www.benchleader.com/benchmarks/arc_agi_3"},{"id":"bfcl_overall","name":"BFCL Overall","category":"agentic","category_label":"Agents & tools","measures":"Berkeley Function Calling Leaderboard overall accuracy.","publisher":"Berkeley Function Calling Leaderboard","publisher_url":"https://gorilla.cs.berkeley.edu/leaderboard.html","models":83,"in_index":true,"index_weight":1,"status":"Active","caveat":"Harness-dependent","leader":"Claude Opus 4.5 at 77.5%","leader_model_url":"https://www.benchleader.com/models/claude-opus-4-5","leader_score":77.47,"unit":"percent","last_result":null,"aliases":["bfcl","function calling","berkeley"],"url":"https://www.benchleader.com/benchmarks/bfcl_overall"},{"id":"tau2_bench","name":"τ²-bench","category":"agentic","category_label":"Agents & tools","measures":"Mean pass^1 across the airline, retail and telecom domains.","publisher":"τ²-bench","publisher_url":"https://taubench.com","models":8,"in_index":true,"index_weight":1,"status":"Active","caveat":"Harness-dependent","leader":"Qwen3.5 397B A17B (thinking) at 87.9%","leader_model_url":"https://www.benchleader.com/models/qwen3-5-397b-a17b","leader_score":87.91,"unit":"percent","last_result":"2026-03-02","aliases":["tau","tau2","tau-bench","taubench","tau bench","tau2 bench","tau2-bench"],"url":"https://www.benchleader.com/benchmarks/tau2_bench"},{"id":"tau3_banking","name":"τ³-bench Banking","category":"agentic","category_label":"Agents & tools","measures":"pass^1 on the banking domain, which Sierra added after the original three. A support agent working a bank's tools against a simulated customer. Sierra's own run; Artificial Analysis' is in the index.","publisher":"τ²-bench","publisher_url":"https://taubench.com","models":26,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"Qwen3.8 Max (max) at 55.1%","leader_model_url":"https://www.benchleader.com/models/qwen3-8-max","leader_score":55.15,"unit":"percent","last_result":"2026-08-04","aliases":[],"url":"https://www.benchleader.com/benchmarks/tau3_banking"},{"id":"aa_critpt","name":"CritPt","category":"reasoning","category_label":"Reasoning","measures":"Unpublished physics research problems at the level of a first PhD project. Run by Artificial Analysis.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/critpt","models":555,"in_index":true,"index_weight":1,"status":"Active","caveat":null,"leader":"Muse Wonder Full (max) at 32.9%","leader_model_url":"https://www.benchleader.com/models/muse-wonder-full","leader_score":32.86,"unit":"percent","last_result":"2026-10-23","aliases":["physics","crit pt"],"url":"https://www.benchleader.com/benchmarks/aa_critpt"},{"id":"aa_gdpval","name":"GDPval-AA v2.1","category":"agentic","category_label":"Agents & tools","measures":"Artificial Analysis' normalised run of OpenAI's GDPval professional-task comparison.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/gdpval-aa","models":167,"in_index":false,"index_weight":null,"status":"Active","caveat":"Judge-based","leader":"Claude Opus 5.5 (max) at 68.3%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":68.31,"unit":"percent","last_result":"2026-10-23","aliases":["gdp val"],"url":"https://www.benchleader.com/benchmarks/aa_gdpval"},{"id":"aa_tau_banking","name":"τ³-Banking (AA)","category":"agentic","category_label":"Agents & tools","measures":"Artificial Analysis' τ²-Bench banking-domain run: a support agent using tools with a simulated customer.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/tau3-banking","models":204,"in_index":true,"index_weight":1,"status":"Active","caveat":"Harness-dependent","leader":"Qwen3.8 Max (max) at 51.3%","leader_model_url":"https://www.benchleader.com/models/qwen3-8-max","leader_score":51.34,"unit":"percent","last_result":"2026-09-11","aliases":["tau","tau2","banking","taubench","tau bench","tau-bench","tau2 bench","tau2-bench"],"url":"https://www.benchleader.com/benchmarks/aa_tau_banking"},{"id":"aa_itbench_sre","name":"ITBench SRE (AA)","category":"agentic","category_label":"Agents & tools","measures":"Site-reliability incident tasks in a live IT environment, run by Artificial Analysis.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/itbench-aa","models":45,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"GPT-5.6 Sol (max) at 56.2%","leader_model_url":"https://www.benchleader.com/models/gpt-5-6-sol","leader_score":56.21,"unit":"percent","last_result":"2026-09-22","aliases":["itbench","sre"],"url":"https://www.benchleader.com/benchmarks/aa_itbench_sre"},{"id":"aa_analyst_agent","name":"Analyst Agent (AA)","category":"agentic","category_label":"Agents & tools","measures":"Multi-step financial-analyst tasks with tools, run by Artificial Analysis.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/aa-analyst-agent","models":50,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"Gemini 3.7 Flash (high) at 60.0%","leader_model_url":"https://www.benchleader.com/models/gemini-3-7-flash","leader_score":60,"unit":"percent","last_result":"2026-10-23","aliases":["analyst"],"url":"https://www.benchleader.com/benchmarks/aa_analyst_agent"},{"id":"aa_apex_agents","name":"APEX-Agents (AA)","category":"agentic","category_label":"Agents & tools","measures":"Artificial Analysis' run of Mercor's long-horizon professional agent tasks.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/apex-agents-aa","models":33,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"Gemini 3.5 Flash (high) at 47.0%","leader_model_url":"https://www.benchleader.com/models/gemini-3-5-flash","leader_score":47.05,"unit":"percent","last_result":"2026-09-18","aliases":["apex","mercor"],"url":"https://www.benchleader.com/benchmarks/aa_apex_agents"},{"id":"vals_math500","name":"MATH 500","category":"math","category_label":"Maths","measures":"Five hundred competition problems in algebra, probability and trigonometry. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/math500","models":57,"in_index":false,"index_weight":null,"status":"Saturated","caveat":"Contamination risk","leader":"Gemini 3 Pro (high) at 96.4%","leader_model_url":"https://www.benchleader.com/models/gemini-3-pro","leader_score":96.4,"unit":"percent","last_result":"2026-01-09","aliases":["math 500"],"url":"https://www.benchleader.com/benchmarks/vals_math500"},{"id":"vals_biomysterybench","name":"BioMysteryBench","category":"knowledge","category_label":"Knowledge","measures":"End-to-end bioinformatics investigations from raw data to a conclusion. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/biomysterybench","models":27,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Claude Sonnet 5.5 at 81.1%","leader_model_url":"https://www.benchleader.com/models/claude-sonnet-5-5","leader_score":81.11,"unit":"percent","last_result":"2026-10-20","aliases":["bio","bioinformatics"],"url":"https://www.benchleader.com/benchmarks/vals_biomysterybench"},{"id":"vals_caselaw","name":"CaseLaw v2","category":"knowledge","category_label":"Knowledge","measures":"Questions over Canadian court cases. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/case_law_v2","models":52,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Grok 4.3 at 79.3%","leader_model_url":"https://www.benchleader.com/models/grok-4-3","leader_score":79.31,"unit":"percent","last_result":"2026-05-04","aliases":["case law","canadian law"],"url":"https://www.benchleader.com/benchmarks/vals_caselaw"},{"id":"vals_claimbench","name":"ClaimBench","category":"knowledge","category_label":"Knowledge","measures":"Judging whether a 10-K excerpt supports a claim, contradicts it or says nothing about it. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/claimbench","models":5,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"GPT-6.1 Sol (max) at 100.0%","leader_model_url":"https://www.benchleader.com/models/gpt-6-1-sol","leader_score":100,"unit":"percent","last_result":"2026-10-23","aliases":[],"url":"https://www.benchleader.com/benchmarks/vals_claimbench"},{"id":"vals_code_migration","name":"Code Migration","category":"coding","category_label":"Coding","measures":"Rewriting whole programs in another language so they still pass the tests. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/code-migration","models":75,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Claude Sonnet 5.5 at 69.8%","leader_model_url":"https://www.benchleader.com/models/claude-sonnet-5-5","leader_score":69.83,"unit":"percent","last_result":"2026-10-18","aliases":["migration","porting"],"url":"https://www.benchleader.com/benchmarks/vals_code_migration"},{"id":"vals_code_migration_11","name":"Code Migration v1.1","category":"coding","category_label":"Coding","measures":"Code Migration v1.1, the version Vals AI now runs new models on. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/code-migration-v1-1","models":26,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Claude Sonnet 5.5 at 86.9%","leader_model_url":"https://www.benchleader.com/models/claude-sonnet-5-5","leader_score":86.88,"unit":"percent","last_result":"2026-10-20","aliases":[],"url":"https://www.benchleader.com/benchmarks/vals_code_migration_11"},{"id":"vals_cua_bench","name":"CUA-bench","category":"agentic","category_label":"Agents & tools","measures":"Playing six commercial video games with only a keyboard and a mouse; half are held out. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/cua_bench","models":9,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"GPT-6 Astra (max) at 19.2%","leader_model_url":"https://www.benchleader.com/models/gpt-6-astra","leader_score":19.17,"unit":"percent","last_result":"2026-10-13","aliases":["computer use","games","cua"],"url":"https://www.benchleader.com/benchmarks/vals_cua_bench"},{"id":"vals_cyber","name":"CyberBench","category":"agentic","category_label":"Agents & tools","measures":"Finding and patching security bugs in real code. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/cyber","models":45,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"Muse Wonder (max) at 80.4%","leader_model_url":"https://www.benchleader.com/models/muse-wonder","leader_score":80.36,"unit":"percent","last_result":"2026-10-12","aliases":["security","cyberbench"],"url":"https://www.benchleader.com/benchmarks/vals_cyber"},{"id":"vals_emb","name":"Excel Modeling Benchmark","category":"knowledge","category_label":"Knowledge","measures":"Building financial models in Excel from a brief. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/emb","models":73,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Claude Fable 5.1 at 76.7%","leader_model_url":"https://www.benchleader.com/models/claude-fable-5-1","leader_score":76.67,"unit":"percent","last_result":"2026-10-20","aliases":["excel","financial modeling"],"url":"https://www.benchleader.com/benchmarks/vals_emb"},{"id":"vals_emb_11","name":"Excel Modeling Benchmark v1.1","category":"knowledge","category_label":"Knowledge","measures":"Excel Modeling Benchmark v1.1, the version Vals AI now runs new models on. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/emb-v1-1","models":31,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Claude Fable 5.1 at 76.7%","leader_model_url":"https://www.benchleader.com/models/claude-fable-5-1","leader_score":76.73,"unit":"percent","last_result":"2026-10-24","aliases":[],"url":"https://www.benchleader.com/benchmarks/vals_emb_11"},{"id":"vals_finance_agent","name":"Finance Agent v2","category":"agentic","category_label":"Agents & tools","measures":"Core financial-analyst tasks with tools. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/fabv2","models":77,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"Gemini 4 Argon (high) at 65.4%","leader_model_url":"https://www.benchleader.com/models/gemini-4-argon","leader_score":65.4,"unit":"percent","last_result":"2026-10-20","aliases":["finance agent","fab"],"url":"https://www.benchleader.com/benchmarks/vals_finance_agent"},{"id":"vals_finance_agent_21","name":"Finance Agent v2.1","category":"agentic","category_label":"Agents & tools","measures":"Finance Agent v2.1, the version Vals AI now runs new models on. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/fabv2-1","models":29,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"Claude Opus 5.5 at 67.8%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":67.85,"unit":"percent","last_result":"2026-10-24","aliases":[],"url":"https://www.benchleader.com/benchmarks/vals_finance_agent_21"},{"id":"vals_finance_index","name":"Vals Finance Index","category":"composite","category_label":"Composite","measures":"Vals AI's composite of its own finance evaluations. Shown for reference.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/finance_index","models":29,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Claude Opus 5.5 at 65.6%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":65.59,"unit":"percent","last_result":"2026-10-24","aliases":[],"url":"https://www.benchleader.com/benchmarks/vals_finance_index"},{"id":"vals_hlab","name":"Harvey's Legal Agent Benchmark v1.1","category":"agentic","category_label":"Agents & tools","measures":"Agentic legal work over files, written with Harvey. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/hlab","models":26,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"Gemini 4 Argon (high) at 27.1%","leader_model_url":"https://www.benchleader.com/models/gemini-4-argon","leader_score":27.08,"unit":"percent","last_result":"2026-10-24","aliases":["harvey","legal agent"],"url":"https://www.benchleader.com/benchmarks/vals_hlab"},{"id":"vals_hlab_v1","name":"Harvey's Legal Agent Benchmark v1","category":"agentic","category_label":"Agents & tools","measures":"The first version of the Harvey legal agent benchmark, which Vals AI keeps online. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/hlab-v1","models":77,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"Muse Spark 1.2 (xhigh) at 25.4%","leader_model_url":"https://www.benchleader.com/models/muse-spark-1-2","leader_score":25.42,"unit":"percent","last_result":"2026-10-18","aliases":[],"url":"https://www.benchleader.com/benchmarks/vals_hlab_v1"},{"id":"vals_legal_research","name":"Legal Research Bench","category":"agentic","category_label":"Agents & tools","measures":"Agentic US legal research questions. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/legal_research","models":81,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"Muse Spark 1.3 (max) at 55.3%","leader_model_url":"https://www.benchleader.com/models/muse-spark-1-3","leader_score":55.29,"unit":"percent","last_result":"2026-10-24","aliases":["legal research"],"url":"https://www.benchleader.com/benchmarks/vals_legal_research"},{"id":"vals_medcode","name":"MedCode","category":"knowledge","category_label":"Knowledge","measures":"Assigning medical billing codes from clinical notes. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/medcode","models":103,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Claude Opus 5 at 63.6%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5","leader_score":63.57,"unit":"percent","last_result":"2026-10-13","aliases":["medical coding","billing"],"url":"https://www.benchleader.com/benchmarks/vals_medcode"},{"id":"vals_medscribe","name":"MedScribe","category":"knowledge","category_label":"Knowledge","measures":"Doctor administrative paperwork from encounter notes. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/medscribe","models":105,"in_index":false,"index_weight":null,"status":"Saturated","caveat":null,"leader":"Claude Opus 5.5 at 91.4%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":91.43,"unit":"percent","last_result":"2026-10-13","aliases":["medical scribe"],"url":"https://www.benchleader.com/benchmarks/vals_medscribe"},{"id":"vals_mmmu_pro","name":"MMMU-Pro (Vals)","category":"multimodal","category_label":"Multimodal","measures":"Vals AI's own run of MMMU-Pro. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/mmmu","models":90,"in_index":false,"index_weight":null,"status":"Near-saturated","caveat":null,"leader":"Claude Fable 5.1 at 90.6%","leader_model_url":"https://www.benchleader.com/models/claude-fable-5-1","leader_score":90.64,"unit":"percent","last_result":"2026-09-01","aliases":["mmmu"],"url":"https://www.benchleader.com/benchmarks/vals_mmmu_pro"},{"id":"vals_mortgage_tax","name":"MortgageTax","category":"multimodal","category_label":"Multimodal","measures":"Reading tax certificate images and extracting the figures. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/mortgage_tax","models":93,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Claude Opus 5 at 72.1%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5","leader_score":72.06,"unit":"percent","last_result":"2026-09-01","aliases":["mortgage","tax certificate"],"url":"https://www.benchleader.com/benchmarks/vals_mortgage_tax"},{"id":"vals_mystery_mechanism","name":"MysteryMechanism","category":"reasoning","category_label":"Reasoning","measures":"Working out a hidden scientific mechanism from sparse observations. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/mysterymechanism","models":28,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"GPT-6 Astra (max) at 53.1%","leader_model_url":"https://www.benchleader.com/models/gpt-6-astra","leader_score":53.15,"unit":"percent","last_result":"2026-10-20","aliases":["mystery mechanism","scientific discovery"],"url":"https://www.benchleader.com/benchmarks/vals_mystery_mechanism"},{"id":"vals_poker","name":"Poker Agent","category":"agentic","category_label":"Agents & tools","measures":"AI agents playing poker against each other. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/poker_agent","models":16,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"GPT-5.2 at 1131.8%","leader_model_url":"https://www.benchleader.com/models/gpt-5-2","leader_score":1131.83,"unit":"percent","last_result":"2025-12-23","aliases":["poker"],"url":"https://www.benchleader.com/benchmarks/vals_poker"},{"id":"vals_programbench","name":"ProgramBench","category":"coding","category_label":"Coding","measures":"Rebuilding programs from compiled binaries. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/programbench","models":61,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Claude Opus 5.5 at 18.5%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":18.5,"unit":"percent","last_result":"2026-10-12","aliases":["reverse engineering","binaries"],"url":"https://www.benchleader.com/benchmarks/vals_programbench"},{"id":"vals_public_benefits","name":"Public Benefits Bench v1.1","category":"knowledge","category_label":"Knowledge","measures":"Navigating SNAP benefits questions for applicants. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/public-benefits-bench","models":48,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Claude Opus 5 at 76.9%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5","leader_score":76.93,"unit":"percent","last_result":"2026-10-12","aliases":["snap","benefits"],"url":"https://www.benchleader.com/benchmarks/vals_public_benefits"},{"id":"vals_robo_bench","name":"Robo-bench","category":"agentic","category_label":"Agents & tools","measures":"Controlling a robot arm and a humanoid in a real-time physics simulation from motor commands and camera images. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/robo_bench","models":7,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"GPT-6.1 Sol (max) at 28.3%","leader_model_url":"https://www.benchleader.com/models/gpt-6-1-sol","leader_score":28.33,"unit":"percent","last_result":"2026-10-13","aliases":[],"url":"https://www.benchleader.com/benchmarks/vals_robo_bench"},{"id":"vals_rsi_index","name":"Vals RSI Index","category":"composite","category_label":"Composite","measures":"Autonomous machine-learning research work, scored against published, human and frontier-model references. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/rsi_index","models":25,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Claude Opus 5.5 (max) at 37.3%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":37.31,"unit":"percent","last_result":"2026-10-15","aliases":[],"url":"https://www.benchleader.com/benchmarks/vals_rsi_index"},{"id":"vals_sage","name":"SAGE","category":"multimodal","category_label":"Multimodal","measures":"Grading handwritten maths work from images. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/sage","models":89,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Claude Opus 4.7 at 56.1%","leader_model_url":"https://www.benchleader.com/models/claude-opus-4-7","leader_score":56.1,"unit":"percent","last_result":"2026-10-08","aliases":["grading","handwritten"],"url":"https://www.benchleader.com/benchmarks/vals_sage"},{"id":"vals_science_index","name":"Vals Science Index","category":"composite","category_label":"Composite","measures":"Vals AI's composite of its biology investigation, mechanism discovery and research workflow evaluations, weighted equally. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/science_index","models":22,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"GPT-6 Astra (max) at 65.1%","leader_model_url":"https://www.benchleader.com/models/gpt-6-astra","leader_score":65.09,"unit":"percent","last_result":"2026-10-21","aliases":[],"url":"https://www.benchleader.com/benchmarks/vals_science_index"},{"id":"vals_skillsbench","name":"SkillsBench","category":"agentic","category_label":"Agents & tools","measures":"How much reusable skills help an agent complete tasks. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/skillsbench","models":33,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"DeepSeek V4.1 Flash (high) at 69.8%","leader_model_url":"https://www.benchleader.com/models/deepseek-v4-1-flash","leader_score":69.8,"unit":"percent","last_result":"2026-09-27","aliases":["skills"],"url":"https://www.benchleader.com/benchmarks/vals_skillsbench"},{"id":"vals_srebench","name":"SREBench","category":"agentic","category_label":"Agents & tools","measures":"Site-reliability incidents to diagnose and fix. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/srebench","models":33,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"GPT-6 Astra (max) at 56.9%","leader_model_url":"https://www.benchleader.com/models/gpt-6-astra","leader_score":56.87,"unit":"percent","last_result":"2026-10-14","aliases":["sre"],"url":"https://www.benchleader.com/benchmarks/vals_srebench"},{"id":"vals_swebench_multimodal","name":"SWE-bench Multimodal","category":"coding","category_label":"Coding","measures":"GitHub issues from JavaScript repositories whose bugs are visual, so fixing them means reading images. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/swebench-multimodal","models":20,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"Claude Opus 5.5 at 70.2%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":70.21,"unit":"percent","last_result":"2026-10-16","aliases":[],"url":"https://www.benchleader.com/benchmarks/vals_swebench_multimodal"},{"id":"vals_tax_agent","name":"Tax Agent Bench","category":"agentic","category_label":"Agents & tools","measures":"Agentic US corporate tax research. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/tax_agent_bench","models":72,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"Qwen3.8 Max Internal at 49.7%","leader_model_url":"https://www.benchleader.com/models/qwen3-8-max-internal","leader_score":49.66,"unit":"percent","last_result":"2026-10-24","aliases":["tax agent"],"url":"https://www.benchleader.com/benchmarks/vals_tax_agent"},{"id":"vals_terminal_bench_2","name":"Terminal-Bench 2.0 (Vals)","category":"agentic","category_label":"Agents & tools","measures":"Terminal-Bench 2.0 in Vals AI's fixed harness. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/terminal-bench-2","models":64,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"GPT-5.5 (high) at 73.2%","leader_model_url":"https://www.benchleader.com/models/gpt-5-5","leader_score":73.2,"unit":"percent","last_result":"2026-06-04","aliases":["terminalbench","tbench"],"url":"https://www.benchleader.com/benchmarks/vals_terminal_bench_2"},{"id":"vals_terminal_bench_4","name":"Terminal-Bench 4.0 (Vals)","category":"agentic","category_label":"Agents & tools","measures":"Terminal-Bench 4.0, the frontier-difficulty set, in Vals AI's harness. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/terminal-bench-4","models":52,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"Claude Opus 5.5 at 65.2%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":65.15,"unit":"percent","last_result":"2026-10-24","aliases":["terminalbench","tbench"],"url":"https://www.benchleader.com/benchmarks/vals_terminal_bench_4"},{"id":"vals_terminal_bench_science","name":"Terminal-Bench Science","category":"agentic","category_label":"Agents & tools","measures":"Expert-authored scientific research workflows in a terminal. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/terminal-bench-science","models":41,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"GPT-6 Astra (max) at 62.9%","leader_model_url":"https://www.benchleader.com/models/gpt-6-astra","leader_score":62.86,"unit":"percent","last_result":"2026-10-21","aliases":["terminalbench science"],"url":"https://www.benchleader.com/benchmarks/vals_terminal_bench_science"},{"id":"vals_time_horizon_ksp","name":"Time Horizon Index: KSP","category":"agentic","category_label":"Agents & tools","measures":"Running a Kerbal Space Program mission end to end. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/time_horizon_index","models":13,"in_index":false,"index_weight":null,"status":"Near-saturated","caveat":"Harness-dependent","leader":"Claude Opus 5.5 at 91.3%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":91.33,"unit":"percent","last_result":"2026-10-07","aliases":["ksp","kerbal"],"url":"https://www.benchleader.com/benchmarks/vals_time_horizon_ksp"},{"id":"vals_multimodal_index","name":"Vals Multimodal Index","category":"composite","category_label":"Composite","measures":"Vals AI's GDP-weighted multimodal composite. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/vals_multimodal_index","models":31,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Claude Fable 5 at 74.2%","leader_model_url":"https://www.benchleader.com/models/claude-fable-5","leader_score":74.15,"unit":"percent","last_result":"2026-08-11","aliases":["multimodal index"],"url":"https://www.benchleader.com/benchmarks/vals_multimodal_index"},{"id":"vals_router_index","name":"Vals Router Index","category":"composite","category_label":"Composite","measures":"Long agentic work in finance, coding, law and tax, where model routers are compared with single models. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/vals_router_index","models":20,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Gemini 4 Argon at 69.2%","leader_model_url":"https://www.benchleader.com/models/gemini-4-argon","leader_score":69.24,"unit":"percent","last_result":"2026-10-20","aliases":[],"url":"https://www.benchleader.com/benchmarks/vals_router_index"},{"id":"vals_vibe_code_bench","name":"Vibe Code Bench 1-100","category":"coding","category_label":"Coding","measures":"Extending a working web app, request by request, one hundred times. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/vcb-1-100","models":25,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Muse Wonder Full (max) at 34.0%","leader_model_url":"https://www.benchleader.com/models/muse-wonder-full","leader_score":33.98,"unit":"percent","last_result":"2026-10-17","aliases":["vibe code","vcb"],"url":"https://www.benchleader.com/benchmarks/vals_vibe_code_bench"},{"id":"vals_vibe_code","name":"Vibe Code Bench v1.1","category":"coding","category_label":"Coding","measures":"Building web apps from scratch from a brief. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/vibe-code","models":115,"in_index":false,"index_weight":null,"status":"Saturated","caveat":"Judge-based","leader":"Claude Sonnet 5.5 at 92.4%","leader_model_url":"https://www.benchleader.com/models/claude-sonnet-5-5","leader_score":92.39,"unit":"percent","last_result":"2026-10-24","aliases":["vibe code"],"url":"https://www.benchleader.com/benchmarks/vals_vibe_code"},{"id":"vals_voice_code_bench","name":"VoiceCodeBench","category":"multimodal","category_label":"Multimodal","measures":"Speech recognition of entities across eight workflow domains. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/voice-code-bench","models":20,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"GPT Live Transcribe at 67.7%","leader_model_url":"https://www.benchleader.com/models/gpt-live-transcribe","leader_score":67.67,"unit":"percent","last_result":"2026-09-24","aliases":["asr","speech","voice"],"url":"https://www.benchleader.com/benchmarks/vals_voice_code_bench"},{"id":"vals_web_search","name":"Web Search Index","category":"agentic","category_label":"Agents & tools","measures":"Answering questions that need live web search, with native and independent tools. Run by Vals AI.","publisher":"Vals AI","publisher_url":"https://www.vals.ai/benchmarks/web_search","models":9,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"GPT-5.6 Sol Keenable (max) at 49.1%","leader_model_url":"https://www.benchleader.com/models/gpt-5-6-sol-keenable","leader_score":49.08,"unit":"percent","last_result":"2026-10-21","aliases":["web search","search"],"url":"https://www.benchleader.com/benchmarks/vals_web_search"},{"id":"scale_audiomc","name":"AudioMC","category":"multimodal","category_label":"Multimodal","measures":"Multiple-choice questions asked and answered in audio. Scale AI.","publisher":"Scale AI SEAL","publisher_url":"https://labs.scale.com/leaderboard/audiomc","models":29,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Gemini 3.8 Flash (high) at 60.4%","leader_model_url":"https://www.benchleader.com/models/gemini-3-8-flash","leader_score":60.4,"unit":"percent","last_result":"2026-10-16","aliases":["audio"],"url":"https://www.benchleader.com/benchmarks/scale_audiomc"},{"id":"scale_audiomc_audio","name":"AudioMC (audio output)","category":"multimodal","category_label":"Multimodal","measures":"AudioMC answered in speech. Scale AI.","publisher":"Scale AI SEAL","publisher_url":"https://labs.scale.com/leaderboard/audiomc-audio","models":15,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"GPT Realtime 2 (xhigh) at 48.5%","leader_model_url":"https://www.benchleader.com/models/gpt-realtime-2","leader_score":48.45,"unit":"percent","last_result":"2026-05-11","aliases":["audio"],"url":"https://www.benchleader.com/benchmarks/scale_audiomc_audio"},{"id":"scale_audiomc_text","name":"AudioMC (text output)","category":"multimodal","category_label":"Multimodal","measures":"AudioMC answered in text. Scale AI.","publisher":"Scale AI SEAL","publisher_url":"https://labs.scale.com/leaderboard/audiomc-text","models":20,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Gemini 3.8 Flash (high) at 60.4%","leader_model_url":"https://www.benchleader.com/models/gemini-3-8-flash","leader_score":60.4,"unit":"percent","last_result":"2026-10-16","aliases":["audio"],"url":"https://www.benchleader.com/benchmarks/scale_audiomc_text"},{"id":"scale_drugdiscovery","name":"DrugDiscoveryBench","category":"knowledge","category_label":"Knowledge","measures":"Drug-discovery reasoning tasks graded by practitioners. Scale AI.","publisher":"Scale AI SEAL","publisher_url":"https://labs.scale.com/leaderboard/drugdiscoverybench","models":19,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"GPT-6 Astra (max) at 68.7%","leader_model_url":"https://www.benchleader.com/models/gpt-6-astra","leader_score":68.7,"unit":"percent","last_result":"2026-10-22","aliases":["drug discovery","pharma"],"url":"https://www.benchleader.com/benchmarks/scale_drugdiscovery"},{"id":"scale_fortress","name":"FORTRESS","category":"safety","category_label":"Safety & honesty","measures":"Robustness to harmful requests without over-refusing benign ones. Scale AI.","publisher":"Scale AI SEAL","publisher_url":"https://labs.scale.com/leaderboard/fortress","models":68,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"DeepSeek R1 at 74.4%","leader_model_url":"https://www.benchleader.com/models/deepseek-r1","leader_score":74.39,"unit":"percent","last_result":"2026-10-19","aliases":["safety","robustness"],"url":"https://www.benchleader.com/benchmarks/scale_fortress"},{"id":"scale_mask","name":"MASK","category":"safety","category_label":"Safety & honesty","measures":"Whether a model states what it believes when pressured to lie. Scale AI.","publisher":"Scale AI SEAL","publisher_url":"https://labs.scale.com/leaderboard/mask","models":64,"in_index":false,"index_weight":null,"status":"Saturated","caveat":null,"leader":"Claude Opus 4.6 (no reasoning) at 96.3%","leader_model_url":"https://www.benchleader.com/models/claude-opus-4-6","leader_score":96.28,"unit":"percent","last_result":"2026-03-25","aliases":["honesty","lying"],"url":"https://www.benchleader.com/benchmarks/scale_mask"},{"id":"scale_propensitybench","name":"PropensityBench","category":"safety","category_label":"Safety & honesty","measures":"How readily an agent takes harmful actions when pressured in agentic settings. Scale AI.","publisher":"Scale AI SEAL","publisher_url":"https://labs.scale.com/leaderboard/propensitybench","models":14,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Gemini 2.5 Pro at 79.0%","leader_model_url":"https://www.benchleader.com/models/gemini-2-5-pro","leader_score":79,"unit":"percent","last_result":"2026-01-15","aliases":["propensity","agentic safety"],"url":"https://www.benchleader.com/benchmarks/scale_propensitybench"},{"id":"scale_scipredict","name":"SciPredict","category":"reasoning","category_label":"Reasoning","measures":"Predicting the outcomes of scientific experiments before they are run. Scale AI.","publisher":"Scale AI SEAL","publisher_url":"https://labs.scale.com/leaderboard/scipredict","models":15,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Gemini 3 Pro at 25.3%","leader_model_url":"https://www.benchleader.com/models/gemini-3-pro","leader_score":25.27,"unit":"percent","last_result":"2026-01-14","aliases":["scipredict","science prediction"],"url":"https://www.benchleader.com/benchmarks/scale_scipredict"},{"id":"scale_swe_bench_pro_private","name":"SWE-Bench Pro (private)","category":"coding","category_label":"Coding","measures":"The held-out commercial-repository set of SWE-Bench Pro. Scale AI.","publisher":"Scale AI SEAL","publisher_url":"https://labs.scale.com/leaderboard/swe_bench_pro_private","models":24,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"Muse Spark 1.1 at 61.5%","leader_model_url":"https://www.benchleader.com/models/muse-spark-1-1","leader_score":61.5,"unit":"percent","last_result":"2026-07-09","aliases":["swe bench pro","swebench pro private"],"url":"https://www.benchleader.com/benchmarks/scale_swe_bench_pro_private"},{"id":"scale_sweatlas_qna","name":"SWE Atlas: Codebase QnA","category":"coding","category_label":"Coding","measures":"Answering questions about a large codebase. Scale AI.","publisher":"Scale AI SEAL","publisher_url":"https://labs.scale.com/leaderboard/sweatlas-qna","models":22,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"Claude Opus 5 (xhigh) at 63.2%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5","leader_score":63.17,"unit":"percent","last_result":"2026-09-09","aliases":["swe atlas","codebase"],"url":"https://www.benchleader.com/benchmarks/scale_sweatlas_qna"},{"id":"scale_sweatlas_refactoring","name":"SWE Atlas: Refactoring","category":"coding","category_label":"Coding","measures":"Refactoring tasks in real repositories. Scale AI.","publisher":"Scale AI SEAL","publisher_url":"https://labs.scale.com/leaderboard/sweatlas-refactoring","models":17,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"GPT-6 Astra (xhigh) at 59.0%","leader_model_url":"https://www.benchleader.com/models/gpt-6-astra","leader_score":59.05,"unit":"percent","last_result":"2026-09-09","aliases":["swe atlas","refactoring"],"url":"https://www.benchleader.com/benchmarks/scale_sweatlas_refactoring"},{"id":"scale_sweatlas_tw","name":"SWE Atlas: Test Writing","category":"coding","category_label":"Coding","measures":"Writing tests for existing code. Scale AI.","publisher":"Scale AI SEAL","publisher_url":"https://labs.scale.com/leaderboard/sweatlas-tw","models":22,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"Claude Fable 5.1 (xhigh) at 67.0%","leader_model_url":"https://www.benchleader.com/models/claude-fable-5-1","leader_score":67.04,"unit":"percent","last_result":"2026-09-09","aliases":["swe atlas","test writing"],"url":"https://www.benchleader.com/benchmarks/scale_sweatlas_tw"},{"id":"scale_vtb","name":"VTB","category":"multimodal","category_label":"Multimodal","measures":"Scale's vision task benchmark across practical image tasks. Scale AI.","publisher":"Scale AI SEAL","publisher_url":"https://labs.scale.com/leaderboard/vtb","models":21,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Muse Spark 1.1 at 44.8%","leader_model_url":"https://www.benchleader.com/models/muse-spark-1-1","leader_score":44.77,"unit":"percent","last_result":"2026-07-09","aliases":["vision task"],"url":"https://www.benchleader.com/benchmarks/scale_vtb"},{"id":"lmarena_text_math","name":"Arena Maths","category":"math","category_label":"Maths","measures":"Text arena rating on maths prompts.","publisher":"Arena (LMArena)","publisher_url":"https://arena.ai/leaderboard/chat/text/math","models":362,"in_index":true,"index_weight":1,"status":"Active","caveat":"Judge-based","leader":"Gemini 4 Argon (high) at 1532","leader_model_url":"https://www.benchleader.com/models/gemini-4-argon","leader_score":1531.7,"unit":"elo","last_result":"2026-10-22","aliases":["arena"],"url":"https://www.benchleader.com/benchmarks/lmarena_text_math"},{"id":"lmarena_text_creative","name":"Arena Creative Writing","category":"human_preference","category_label":"Human preference","measures":"Text arena rating on creative-writing prompts.","publisher":"Arena (LMArena)","publisher_url":"https://arena.ai/leaderboard/chat/text/creative-writing","models":371,"in_index":false,"index_weight":null,"status":"Active","caveat":"Judge-based","leader":"Claude Opus 5.5 (high) at 1520","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":1519.6,"unit":"elo","last_result":"2026-10-22","aliases":["writing","creative","arena"],"url":"https://www.benchleader.com/benchmarks/lmarena_text_creative"},{"id":"lmarena_text_if","name":"Arena Instruction Following","category":"instruction","category_label":"Instruction following","measures":"Text arena rating on prompts with explicit instructions.","publisher":"Arena (LMArena)","publisher_url":"https://arena.ai/leaderboard/chat/text/instruction-following","models":373,"in_index":false,"index_weight":null,"status":"Active","caveat":"Judge-based","leader":"Claude Opus 5.5 (high) at 1521","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":1521.4,"unit":"elo","last_result":"2026-10-22","aliases":["instruction following","arena"],"url":"https://www.benchleader.com/benchmarks/lmarena_text_if"},{"id":"lmarena_text_multiturn","name":"Arena Multi-turn","category":"human_preference","category_label":"Human preference","measures":"Text arena rating on conversations with more than one turn.","publisher":"Arena (LMArena)","publisher_url":"https://arena.ai/leaderboard/chat/text/multi-turn","models":371,"in_index":false,"index_weight":null,"status":"Active","caveat":"Judge-based","leader":"Gemini 4 Argon (high) at 1544","leader_model_url":"https://www.benchleader.com/models/gemini-4-argon","leader_score":1543.8,"unit":"elo","last_result":"2026-10-22","aliases":["multi turn","conversation","arena"],"url":"https://www.benchleader.com/benchmarks/lmarena_text_multiturn"},{"id":"lmarena_text_longer","name":"Arena Longer Queries","category":"human_preference","category_label":"Human preference","measures":"Text arena rating on long prompts.","publisher":"Arena (LMArena)","publisher_url":"https://arena.ai/leaderboard/chat/text/longer-query","models":352,"in_index":false,"index_weight":null,"status":"Active","caveat":"Judge-based","leader":"Gemini 4 Argon (high) at 1535","leader_model_url":"https://www.benchleader.com/models/gemini-4-argon","leader_score":1534.5,"unit":"elo","last_result":"2026-10-22","aliases":["long query","arena"],"url":"https://www.benchleader.com/benchmarks/lmarena_text_longer"},{"id":"lmarena_search","name":"Arena Search","category":"agentic","category_label":"Agents & tools","measures":"Pairwise votes on answers that used web search.","publisher":"Arena (LMArena)","publisher_url":"https://arena.ai/leaderboard/chat/search","models":41,"in_index":false,"index_weight":null,"status":"Active","caveat":"Judge-based","leader":"GPT-5.6 Sol (xhigh) at 1258","leader_model_url":"https://www.benchleader.com/models/gpt-5-6-sol","leader_score":1257.9,"unit":"elo","last_result":"2026-10-13","aliases":["search arena","web search","arena"],"url":"https://www.benchleader.com/benchmarks/lmarena_search"},{"id":"lmarena_document","name":"Arena Document","category":"long_context","category_label":"Long context","measures":"Pairwise votes on answers about uploaded documents.","publisher":"Arena (LMArena)","publisher_url":"https://arena.ai/leaderboard/chat/document","models":44,"in_index":false,"index_weight":null,"status":"Active","caveat":"Judge-based","leader":"Claude Fable 5 at 1499","leader_model_url":"https://www.benchleader.com/models/claude-fable-5","leader_score":1498.9,"unit":"elo","last_result":"2026-09-13","aliases":["document arena","arena"],"url":"https://www.benchleader.com/benchmarks/lmarena_document"},{"id":"chess_puzzles","name":"Chess Puzzles","category":"reasoning","category_label":"Reasoning","measures":"Solving chess puzzles from a board description. Run by Epoch AI.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/chess-puzzles","models":146,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"GPT-6 Astra Pro (max) at 76.0%","leader_model_url":"https://www.benchleader.com/models/gpt-6-astra-pro","leader_score":76,"unit":"percent","last_result":"2026-10-17","aliases":["chess"],"url":"https://www.benchleader.com/benchmarks/chess_puzzles"},{"id":"ebr_bench","name":"EBR-bench","category":"knowledge","category_label":"Knowledge","measures":"Evidence-based reasoning over research literature. Run by Epoch AI.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/ebr-bench","models":27,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"GPT-6 Astra (max) at 76.2%","leader_model_url":"https://www.benchleader.com/models/gpt-6-astra","leader_score":76.19,"unit":"percent","last_result":"2026-10-12","aliases":["ebr"],"url":"https://www.benchleader.com/benchmarks/ebr_bench"},{"id":"mystery_game_puzzles","name":"Mystery Game Puzzles","category":"reasoning","category_label":"Reasoning","measures":"Deduction puzzles in the style of mystery games. Run by Epoch AI.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/mystery-game-puzzles","models":141,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"GPT-6 Astra (max) at 84.0%","leader_model_url":"https://www.benchleader.com/models/gpt-6-astra","leader_score":84,"unit":"percent","last_result":"2026-10-17","aliases":["mystery","deduction"],"url":"https://www.benchleader.com/benchmarks/mystery_game_puzzles"},{"id":"mirrorcode","name":"MirrorCode","category":"coding","category_label":"Coding","measures":"Reproducing a program's behaviour from its outputs. Run by Epoch AI.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/mirrorcode","models":10,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Claude Opus 5.5 (max) at 77.4%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":77.4,"unit":"percent","last_result":"2026-09-30","aliases":["mirror code"],"url":"https://www.benchleader.com/benchmarks/mirrorcode"},{"id":"balrog","name":"BALROG","category":"agentic","category_label":"Agents & tools","measures":"Progress through classic text and grid games such as NetHack.","publisher":"BALROG","publisher_url":"https://balrogai.com","models":37,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"GPT-6 Astra (max) at 68.3%","leader_model_url":"https://www.benchleader.com/models/gpt-6-astra","leader_score":68.3,"unit":"percent","last_result":"2026-09-03","aliases":["balrog","nethack","games"],"url":"https://www.benchleader.com/benchmarks/balrog"},{"id":"cad_eval","name":"CadEval","category":"coding","category_label":"Coding","measures":"Generating CAD models from specifications.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/cad-eval","models":15,"in_index":false,"index_weight":null,"status":"Retired","caveat":null,"leader":"o3 (medium) at 74.0%","leader_model_url":"https://www.benchleader.com/models/o3","leader_score":74,"unit":"percent","last_result":"2025-04-16","aliases":["cad"],"url":"https://www.benchleader.com/benchmarks/cad_eval"},{"id":"deepresearch_bench","name":"DeepResearch Bench","category":"agentic","category_label":"Agents & tools","measures":"Long-form research reports judged for depth and accuracy.","publisher":"DeepResearch Bench","publisher_url":"https://deepresearch-bench.github.io","models":40,"in_index":false,"index_weight":null,"status":"Active","caveat":"Judge-based","leader":"Claude Opus 4.6 (high) at 55.3%","leader_model_url":"https://www.benchleader.com/models/claude-opus-4-6","leader_score":55.3,"unit":"percent","last_result":"2026-05-28","aliases":["deep research"],"url":"https://www.benchleader.com/benchmarks/deepresearch_bench"},{"id":"geobench","name":"GeoBench","category":"multimodal","category_label":"Multimodal","measures":"Identifying where a photo was taken.","publisher":"GeoBench","publisher_url":"https://geobench.org","models":27,"in_index":false,"index_weight":null,"status":"Near-saturated","caveat":null,"leader":"Gemini 3 Flash at 88.0%","leader_model_url":"https://www.benchleader.com/models/gemini-3-flash","leader_score":88,"unit":"percent","last_result":"2025-12-17","aliases":["geo","geoguessr"],"url":"https://www.benchleader.com/benchmarks/geobench"},{"id":"lech_mazur_writing","name":"Lech Mazur Creative Writing","category":"human_preference","category_label":"Human preference","measures":"Short-story writing judged by a panel of models.","publisher":"Lech Mazur","publisher_url":"https://github.com/lechmazur/writing","models":42,"in_index":false,"index_weight":null,"status":"Retired","caveat":"Judge-based","leader":"GPT-5 (medium) at 86.0%","leader_model_url":"https://www.benchleader.com/models/gpt-5","leader_score":86,"unit":"percent","last_result":"2025-08-07","aliases":["creative writing","lech mazur"],"url":"https://www.benchleader.com/benchmarks/lech_mazur_writing"},{"id":"the_agent_company","name":"TheAgentCompany","category":"agentic","category_label":"Agents & tools","measures":"Office tasks in a simulated software company.","publisher":"TheAgentCompany","publisher_url":"https://the-agent-company.com","models":14,"in_index":false,"index_weight":null,"status":"Retired","caveat":"Harness-dependent","leader":"DeepSeek V3.2 at 42.9%","leader_model_url":"https://www.benchleader.com/models/deepseek-v3-2","leader_score":42.9,"unit":"percent","last_result":"2025-09-29","aliases":["agent company"],"url":"https://www.benchleader.com/benchmarks/the_agent_company"},{"id":"vpct","name":"VPCT","category":"multimodal","category_label":"Multimodal","measures":"Visual physics comprehension: predicting what happens next in a scene.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/vpct","models":30,"in_index":false,"index_weight":null,"status":"Near-saturated","caveat":null,"leader":"Gemini 3 Pro at 91.0%","leader_model_url":"https://www.benchleader.com/models/gemini-3-pro","leader_score":91,"unit":"percent","last_result":"2025-12-17","aliases":["visual physics"],"url":"https://www.benchleader.com/benchmarks/vpct"},{"id":"posttrainbench","name":"PostTrainBench","category":"agentic","category_label":"Agents & tools","measures":"Post-training a small model to a target, end to end.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/post-train-bench","models":12,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"Claude Fable 5 (max) at 41.8%","leader_model_url":"https://www.benchleader.com/models/claude-fable-5","leader_score":41.79,"unit":"percent","last_result":"2026-07-24","aliases":["post training"],"url":"https://www.benchleader.com/benchmarks/posttrainbench"},{"id":"exploitbench","name":"ExploitBench","category":"agentic","category_label":"Agents & tools","measures":"Developing working exploits for known vulnerabilities.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/exploitbench","models":9,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"Claude Mythos Preview at 73.8%","leader_model_url":"https://www.benchleader.com/models/claude-mythos-early","leader_score":73.8,"unit":"percent","last_result":"2026-04-23","aliases":["exploit"],"url":"https://www.benchleader.com/benchmarks/exploitbench"},{"id":"cl_bench","name":"CL-bench","category":"reasoning","category_label":"Reasoning","measures":"Continual-learning style tasks that require using information given in context.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/cl-bench","models":21,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"GPT-5.4 (xhigh) at 27.9%","leader_model_url":"https://www.benchleader.com/models/gpt-5-4","leader_score":27.9,"unit":"percent","last_result":"2026-03-31","aliases":["cl bench","continual learning"],"url":"https://www.benchleader.com/benchmarks/cl_bench"},{"id":"cl_bench_life","name":"CL-bench Life","category":"reasoning","category_label":"Reasoning","measures":"The everyday-life subset of CL-bench.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/cl-bench-life","models":16,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"GPT-5.5 (high) at 22.2%","leader_model_url":"https://www.benchleader.com/models/gpt-5-5","leader_score":22.2,"unit":"percent","last_result":"2026-04-24","aliases":["cl bench"],"url":"https://www.benchleader.com/benchmarks/cl_bench_life"},{"id":"surface_evolver_bench","name":"Surface Evolver Bench","category":"coding","category_label":"Coding","measures":"Scientific computing with the Surface Evolver package.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/surface-evolver-bench","models":29,"in_index":false,"index_weight":null,"status":"Near-saturated","caveat":null,"leader":"Claude Fable 5 (high) at 95.0%","leader_model_url":"https://www.benchleader.com/models/claude-fable-5","leader_score":95,"unit":"percent","last_result":"2026-10-06","aliases":["surface evolver"],"url":"https://www.benchleader.com/benchmarks/surface_evolver_bench"},{"id":"metr_time_horizons","name":"METR Time Horizons","category":"agentic","category_label":"Agents & tools","measures":"METR's task suite behind the 50%-time-horizon measure.","publisher":"METR","publisher_url":"https://metr.org","models":40,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"Claude Mythos Preview at 85.2%","leader_model_url":"https://www.benchleader.com/models/claude-mythos-early","leader_score":85.21,"unit":"percent","last_result":"2026-04-07","aliases":["metr","time horizon"],"url":"https://www.benchleader.com/benchmarks/metr_time_horizons"},{"id":"deepswe","name":"DeepSWE v1.1","category":"coding","category_label":"Coding","measures":"Software-engineering tasks in an agent harness with reasoning effort recorded.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/deepswe","models":69,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"GPT-6 Astra (xhigh) at 74.1%","leader_model_url":"https://www.benchleader.com/models/gpt-6-astra","leader_score":74.12,"unit":"percent","last_result":"2026-09-03","aliases":["deep swe"],"url":"https://www.benchleader.com/benchmarks/deepswe"},{"id":"lmca","name":"LMCA","category":"reasoning","category_label":"Reasoning","measures":"A capability assessment with a published score ceiling.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/lmca","models":192,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Claude Sonnet 5.5 (max) at 68.3%","leader_model_url":"https://www.benchleader.com/models/claude-sonnet-5-5","leader_score":68.26,"unit":"percent","last_result":"2026-10-12","aliases":["lmca"],"url":"https://www.benchleader.com/benchmarks/lmca"},{"id":"dtbench","name":"DTBench","category":"reasoning","category_label":"Reasoning","measures":"Decision-theory dilemmas scored against expert answers.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/dtbench","models":229,"in_index":false,"index_weight":null,"status":"Saturated","caveat":null,"leader":"Claude Opus 5.5 (max) at 98.9%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":98.93,"unit":"percent","last_result":"2026-10-12","aliases":["decision theory"],"url":"https://www.benchleader.com/benchmarks/dtbench"},{"id":"forecastbench","name":"ForecastBench","category":"reasoning","category_label":"Reasoning","measures":"Forecasting real future events, scored as they resolve.","publisher":"ForecastBench","publisher_url":"https://www.forecastbench.org","models":85,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"GPT-5.5 (high) at 67.0%","leader_model_url":"https://www.benchleader.com/models/gpt-5-5","leader_score":67,"unit":"percent","last_result":"2026-08-13","aliases":["forecasting"],"url":"https://www.benchleader.com/benchmarks/forecastbench"},{"id":"video_mme","name":"Video-MME","category":"multimodal","category_label":"Multimodal","measures":"Video understanding across durations, without subtitles.","publisher":"Video-MME","publisher_url":"https://video-mme.github.io","models":42,"in_index":false,"index_weight":null,"status":"Retired","caveat":null,"leader":"video-SALMONN 2+ at 79.7%","leader_model_url":"https://www.benchleader.com/models/video-salmonn-2plus","leader_score":79.7,"unit":"percent","last_result":"2025-06-18","aliases":["video"],"url":"https://www.benchleader.com/benchmarks/video_mme"},{"id":"cursorbench","name":"CursorBench","category":"coding","category_label":"Coding","measures":"Cursor's internal coding-agent evaluation, with reasoning level and cost.","publisher":"Cursor","publisher_url":"https://cursor.com","models":78,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"Claude Opus 5.5 (max) at 57.8%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":57.8,"unit":"percent","last_result":"2026-10-12","aliases":["cursor"],"url":"https://www.benchleader.com/benchmarks/cursorbench"},{"id":"gbaeval","name":"GBAEval","category":"agentic","category_label":"Agents & tools","measures":"Playing Game Boy Advance games as an agent.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/gbaeval","models":23,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"Claude Opus 5 at 79.6%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5","leader_score":79.6,"unit":"percent","last_result":"2026-07-24","aliases":["gba","game boy"],"url":"https://www.benchleader.com/benchmarks/gbaeval"},{"id":"ale_bench","name":"ALE-Bench","category":"coding","category_label":"Coding","measures":"Long-horizon algorithm engineering on AtCoder heuristic contests.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/ale-bench","models":126,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"GPT-6 Astra (max) at 2951.3","leader_model_url":"https://www.benchleader.com/models/gpt-6-astra","leader_score":2951.32,"unit":"score","last_result":"2026-10-15","aliases":["atcoder","heuristic"],"url":"https://www.benchleader.com/benchmarks/ale_bench"},{"id":"algotune","name":"AlgoTune","category":"coding","category_label":"Coding","measures":"Speeding up numeric code beyond reference implementations.","publisher":"AlgoTune","publisher_url":"https://algotune.io","models":18,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"GPT-5.2 (medium) at 2.0","leader_model_url":"https://www.benchleader.com/models/gpt-5-2","leader_score":2.05,"unit":"score","last_result":"2026-03-05","aliases":["algotune","optimisation"],"url":"https://www.benchleader.com/benchmarks/algotune"},{"id":"vending_bench_2","name":"Vending-Bench 2","category":"agentic","category_label":"Agents & tools","measures":"Running a simulated vending-machine business for a year.","publisher":"Andon Labs","publisher_url":"https://andonlabs.com","models":69,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"GPT-6 Astra at 15514.7","leader_model_url":"https://www.benchleader.com/models/gpt-6-astra","leader_score":15514.7,"unit":"score","last_result":"2026-10-12","aliases":["vending","andon"],"url":"https://www.benchleader.com/benchmarks/vending_bench_2"},{"id":"blueprint_bench_2","name":"Blueprint-Bench 2","category":"multimodal","category_label":"Multimodal","measures":"Turning photos of rooms into floor plans.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/blueprint-bench-2","models":33,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Gemini 4 Argon at 54.5%","leader_model_url":"https://www.benchleader.com/models/gemini-4-argon","leader_score":54.45,"unit":"percent","last_result":"2026-10-07","aliases":["blueprint","floor plan"],"url":"https://www.benchleader.com/benchmarks/blueprint_bench_2"},{"id":"mindcube","name":"MindCube","category":"multimodal","category_label":"Multimodal","measures":"Spatial reasoning from multiple views.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/mindcube","models":5,"in_index":false,"index_weight":null,"status":"Retired","caveat":null,"leader":"Gemma 3 12B at 46.7%","leader_model_url":"https://www.benchleader.com/models/gemma-3-12b","leader_score":46.67,"unit":"percent","last_result":"2025-05-22","aliases":["spatial"],"url":"https://www.benchleader.com/benchmarks/mindcube"},{"id":"spatialviz_bench","name":"SpatialViz-Bench","category":"multimodal","category_label":"Multimodal","measures":"Spatial visualisation: rotation, folding and animation.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/spatialviz-bench","models":8,"in_index":false,"index_weight":null,"status":"Retired","caveat":null,"leader":"Gemini 2.5 Pro at 44.7%","leader_model_url":"https://www.benchleader.com/models/gemini-2-5-pro","leader_score":44.66,"unit":"percent","last_result":"2025-06-17","aliases":["spatial","rotation"],"url":"https://www.benchleader.com/benchmarks/spatialviz_bench"},{"id":"gdp_pdf","name":"GDP.pdf","category":"knowledge","category_label":"Knowledge","measures":"Professional document tasks over PDFs.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/gdp-pdf","models":56,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"GPT-6.1 Sol (max) at 35.0%","leader_model_url":"https://www.benchleader.com/models/gpt-6-1-sol","leader_score":35,"unit":"percent","last_result":"2026-10-12","aliases":["gdp pdf","pdf"],"url":"https://www.benchleader.com/benchmarks/gdp_pdf"},{"id":"frontierswe","name":"FrontierSWE","category":"coding","category_label":"Coding","measures":"Frontier-difficulty software-engineering tasks across implementation, performance and research.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/frontierswe","models":19,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"GPT-6 Astra (max) at 65.5%","leader_model_url":"https://www.benchleader.com/models/gpt-6-astra","leader_score":65.51,"unit":"percent","last_result":"2026-10-07","aliases":["frontier swe"],"url":"https://www.benchleader.com/benchmarks/frontierswe"},{"id":"btf3","name":"BTF-3","category":"reasoning","category_label":"Reasoning","measures":"Forecasting tournament pooled score.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/btf-3","models":13,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"GLM 5.3 Flash at 14.9%","leader_model_url":"https://www.benchleader.com/models/glm-5-3-flash","leader_score":14.9,"unit":"percent","last_result":"2026-09-03","aliases":["btf","forecasting"],"url":"https://www.benchleader.com/benchmarks/btf3"},{"id":"mmlu","name":"MMLU","category":"knowledge","category_label":"Knowledge","measures":"Fifty-seven-subject multiple-choice exam of world knowledge. A 2020 benchmark, saturated for current models; kept for history.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/mmlu","models":120,"in_index":false,"index_weight":null,"status":"Retired","caveat":"Contamination risk","leader":"GPT-4o at 88.1%","leader_model_url":"https://www.benchleader.com/models/gpt-4o","leader_score":88.1,"unit":"percent","last_result":"2024-12-26","aliases":["mmlu"],"url":"https://www.benchleader.com/benchmarks/mmlu"},{"id":"gsm8k","name":"GSM8K","category":"math","category_label":"Maths","measures":"Grade-school maths word problems. A 2021 benchmark, saturated for current models; kept for history.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/gsm8k","models":73,"in_index":false,"index_weight":null,"status":"Retired","caveat":"Contamination risk","leader":"Deepseek Coder v2 at 94.5%","leader_model_url":"https://www.benchleader.com/models/deepseek-coder-v2","leader_score":94.5,"unit":"percent","last_result":"2024-11-29","aliases":["gsm8k","grade school"],"url":"https://www.benchleader.com/benchmarks/gsm8k"},{"id":"hellaswag","name":"HellaSwag","category":"reasoning","category_label":"Reasoning","measures":"Commonsense sentence completion. A 2019 benchmark, saturated for current models; kept for history.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/hella-swag","models":68,"in_index":false,"index_weight":null,"status":"Retired","caveat":"Contamination risk","leader":"GPT-4 at 95.3%","leader_model_url":"https://www.benchleader.com/models/gpt-4","leader_score":95.3,"unit":"percent","last_result":"2024-12-26","aliases":[],"url":"https://www.benchleader.com/benchmarks/hellaswag"},{"id":"bbh","name":"BIG-Bench Hard","category":"reasoning","category_label":"Reasoning","measures":"Twenty-three hard tasks from BIG-Bench. A 2022 benchmark, saturated for current models; kept for history.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/bbh","models":43,"in_index":false,"index_weight":null,"status":"Retired","caveat":"Contamination risk","leader":"Gemini 1.5 Pro 001 at 89.2%","leader_model_url":"https://www.benchleader.com/models/gemini-1-5-pro-001","leader_score":89.2,"unit":"percent","last_result":"2024-12-26","aliases":["bbh","big bench"],"url":"https://www.benchleader.com/benchmarks/bbh"},{"id":"winogrande","name":"WinoGrande","category":"reasoning","category_label":"Reasoning","measures":"Pronoun-resolution commonsense questions. A 2019 benchmark, saturated for current models; kept for history.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/wino-grande","models":74,"in_index":false,"index_weight":null,"status":"Retired","caveat":"Contamination risk","leader":"Llama 3.1 405B at 89.2%","leader_model_url":"https://www.benchleader.com/models/llama-3-1-405b","leader_score":89.2,"unit":"percent","last_result":"2024-12-26","aliases":[],"url":"https://www.benchleader.com/benchmarks/winogrande"},{"id":"triviaqa","name":"TriviaQA","category":"knowledge","category_label":"Knowledge","measures":"Trivia questions with evidence documents. A 2017 benchmark, saturated for current models; kept for history.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/trivia-qa","models":36,"in_index":false,"index_weight":null,"status":"Retired","caveat":"Contamination risk","leader":"Llama 2-70B at 87.6%","leader_model_url":"https://www.benchleader.com/models/llama-2-70b","leader_score":87.6,"unit":"percent","last_result":"2024-12-26","aliases":["trivia"],"url":"https://www.benchleader.com/benchmarks/triviaqa"},{"id":"arc_ai2","name":"ARC (AI2 Challenge)","category":"reasoning","category_label":"Reasoning","measures":"Grade-school science questions, challenge set. A 2018 benchmark, saturated for current models; kept for history.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/arc-ai2","models":77,"in_index":false,"index_weight":null,"status":"Retired","caveat":"Contamination risk","leader":"DeepSeek V3 at 95.3%","leader_model_url":"https://www.benchleader.com/models/deepseek-v3","leader_score":95.3,"unit":"percent","last_result":"2024-12-26","aliases":["ai2 arc"],"url":"https://www.benchleader.com/benchmarks/arc_ai2"},{"id":"piqa","name":"PIQA","category":"reasoning","category_label":"Reasoning","measures":"Physical commonsense questions. A 2019 benchmark, saturated for current models; kept for history.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/piqa","models":56,"in_index":false,"index_weight":null,"status":"Retired","caveat":"Contamination risk","leader":"GPT-4o mini at 88.7%","leader_model_url":"https://www.benchleader.com/models/gpt-4o-mini","leader_score":88.7,"unit":"percent","last_result":"2024-12-26","aliases":[],"url":"https://www.benchleader.com/benchmarks/piqa"},{"id":"openbookqa","name":"OpenBookQA","category":"knowledge","category_label":"Knowledge","measures":"Elementary science questions with an open book of facts. A 2018 benchmark, saturated for current models; kept for history.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/open-book-qa","models":41,"in_index":false,"index_weight":null,"status":"Retired","caveat":"Contamination risk","leader":"Phi 3 Mini at 88.0%","leader_model_url":"https://www.benchleader.com/models/phi-3-mini","leader_score":88,"unit":"percent","last_result":"2024-04-23","aliases":[],"url":"https://www.benchleader.com/benchmarks/openbookqa"},{"id":"superglue","name":"SuperGLUE","category":"reasoning","category_label":"Reasoning","measures":"A suite of language-understanding tasks. A 2019 benchmark, saturated for current models; kept for history.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/superglue","models":8,"in_index":false,"index_weight":null,"status":"Retired","caveat":"Contamination risk","leader":"T5-11B at 88.9%","leader_model_url":"https://www.benchleader.com/models/t5-11b","leader_score":88.9,"unit":"percent","last_result":"2022-01-27","aliases":[],"url":"https://www.benchleader.com/benchmarks/superglue"},{"id":"lambada","name":"LAMBADA","category":"knowledge","category_label":"Knowledge","measures":"Predicting the last word of a passage. A 2016 benchmark, saturated for current models; kept for history.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/lambada","models":24,"in_index":false,"index_weight":null,"status":"Retired","caveat":"Contamination risk","leader":"Megatron-Turing NLG 530B at 87.2%","leader_model_url":"https://www.benchleader.com/models/megatron-turing-nlg-530b","leader_score":87.15,"unit":"percent","last_result":"2023-11-30","aliases":[],"url":"https://www.benchleader.com/benchmarks/lambada"},{"id":"scienceqa","name":"ScienceQA","category":"multimodal","category_label":"Multimodal","measures":"Science questions with diagrams. A 2022 benchmark, saturated for current models; kept for history.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/science-qa","models":23,"in_index":false,"index_weight":null,"status":"Retired","caveat":"Contamination risk","leader":"Phi 3.5 Vision at 91.3%","leader_model_url":"https://www.benchleader.com/models/phi-3-5-vision","leader_score":91.3,"unit":"percent","last_result":"2024-08-16","aliases":[],"url":"https://www.benchleader.com/benchmarks/scienceqa"},{"id":"anli","name":"ANLI","category":"reasoning","category_label":"Reasoning","measures":"Adversarially written natural-language inference. A 2019 benchmark, saturated for current models; kept for history.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks/adversarial-nli","models":11,"in_index":false,"index_weight":null,"status":"Retired","caveat":"Contamination risk","leader":"GPT 3.5 Turbo 1106 at 58.1%","leader_model_url":"https://www.benchleader.com/models/gpt-3-5-turbo-1106","leader_score":58.1,"unit":"percent","last_result":"2024-04-23","aliases":["nli"],"url":"https://www.benchleader.com/benchmarks/anli"},{"id":"csqa2","name":"CommonsenseQA 2","category":"reasoning","category_label":"Reasoning","measures":"Commonsense yes/no questions. A 2022 benchmark, saturated for current models; kept for history.","publisher":"Epoch AI Benchmarking Hub","publisher_url":"https://epoch.ai/benchmarks","models":6,"in_index":false,"index_weight":null,"status":"Retired","caveat":"Contamination risk","leader":"T5-11B at 67.8%","leader_model_url":"https://www.benchleader.com/models/t5-11b","leader_score":67.8,"unit":"percent","last_result":"2023-07-18","aliases":["commonsense"],"url":"https://www.benchleader.com/benchmarks/csqa2"},{"id":"aa_terminalbench_40","name":"Terminal-Bench 4.0 (AA)","category":"coding","category_label":"Coding","measures":"Real command-line tasks in a sandboxed terminal, version 4.0. Run by Artificial Analysis; the Terminal-Bench team's own run is the one in the index.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/terminalbench-4-0","models":221,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"Kimi K3.1 Pro at 67.2%","leader_model_url":"https://www.benchleader.com/models/kimi-k3-1-pro","leader_score":67.17,"unit":"percent","last_result":"2026-10-23","aliases":[],"url":"https://www.benchleader.com/benchmarks/aa_terminalbench_40"},{"id":"aa_terminalbench_21","name":"Terminal-Bench 2.1 (AA)","category":"coding","category_label":"Coding","measures":"The previous Terminal-Bench release, kept because far more models have been run on it than on 4.0.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/terminalbench-2-1","models":236,"in_index":false,"index_weight":null,"status":"Near-saturated","caveat":"Harness-dependent","leader":"Claude Fable 5.1 (max) at 91.4%","leader_model_url":"https://www.benchleader.com/models/claude-fable-5-1","leader_score":91.39,"unit":"percent","last_result":"2026-10-15","aliases":[],"url":"https://www.benchleader.com/benchmarks/aa_terminalbench_21"},{"id":"aa_automationbench","name":"AutomationBench","category":"agentic","category_label":"Agents & tools","measures":"Office automation tasks run end to end. Scored on partial credit, so a model that gets most of a workflow right still registers.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/automationbench-aa","models":224,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"Gemini 4 Argon (high) at 77.5%","leader_model_url":"https://www.benchleader.com/models/gemini-4-argon","leader_score":77.51,"unit":"percent","last_result":"2026-10-23","aliases":[],"url":"https://www.benchleader.com/benchmarks/aa_automationbench"},{"id":"aa_gdp_pdf","name":"GDP.pdf (AA)","category":"agentic","category_label":"Agents & tools","measures":"Document work on real PDFs, scored all-or-nothing: every required element has to be right for the task to pass, which is why the scores look low.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/gdp-pdf","models":223,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"Muse Wonder Full (max) at 32.6%","leader_model_url":"https://www.benchleader.com/models/muse-wonder-full","leader_score":32.6,"unit":"percent","last_result":"2026-10-23","aliases":[],"url":"https://www.benchleader.com/benchmarks/aa_gdp_pdf"},{"id":"aa_mlcr","name":"MLCR","category":"long_context","category_label":"Long context","measures":"Medical long-context reasoning: clinical questions answered over long patient records, scored on accuracy, completeness and concision together.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/mlcr-aa","models":170,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"GPT-6 Astra (max) at 41.7%","leader_model_url":"https://www.benchleader.com/models/gpt-6-astra","leader_score":41.67,"unit":"percent","last_result":"2026-10-19","aliases":[],"url":"https://www.benchleader.com/benchmarks/aa_mlcr"},{"id":"aa_harvey_lab","name":"Harvey LAB-AA v1.1","category":"agentic","category_label":"Agents & tools","measures":"Legal work drawn from practice, built with Harvey. Run by Artificial Analysis.","publisher":"Harvey","publisher_url":"https://www.harvey.ai","models":27,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"Muse Wonder Full (max) at 13.1%","leader_model_url":"https://www.benchleader.com/models/muse-wonder-full","leader_score":13.06,"unit":"percent","last_result":"2026-10-12","aliases":[],"url":"https://www.benchleader.com/benchmarks/aa_harvey_lab"},{"id":"aa_enterprise_ops_gym","name":"EnterpriseOps-Gym","category":"agentic","category_label":"Agents & tools","measures":"Enterprise operations tasks in a simulated company: tickets, approvals and systems a model has to drive to finish a job.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/enterprise-ops-gym-aa","models":49,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"Claude Fable 5 (max) at 51.1%","leader_model_url":"https://www.benchleader.com/models/claude-fable-5","leader_score":51.12,"unit":"percent","last_result":"2026-09-18","aliases":[],"url":"https://www.benchleader.com/benchmarks/aa_enterprise_ops_gym"},{"id":"aa_omniscience_accuracy","name":"AA-Omniscience: accuracy","category":"knowledge","category_label":"Knowledge","measures":"The share of AA-Omniscience questions answered correctly, before any penalty for confident wrong answers.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/omniscience","models":552,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Claude Fable 5.1 (max) at 67.2%","leader_model_url":"https://www.benchleader.com/models/claude-fable-5-1","leader_score":67.23,"unit":"percent","last_result":"2026-10-23","aliases":[],"url":"https://www.benchleader.com/benchmarks/aa_omniscience_accuracy"},{"id":"aa_omniscience_non_hallucination","name":"AA-Omniscience: non-hallucination","category":"knowledge","category_label":"Knowledge","measures":"How often a model declines to answer rather than inventing one, on questions it does not know. The counterweight to the accuracy score.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/omniscience","models":552,"in_index":false,"index_weight":null,"status":"Near-saturated","caveat":null,"leader":"MiniCPM5-1B (no reasoning) at 99.1%","leader_model_url":"https://www.benchleader.com/models/minicpm5-1b","leader_score":99.1,"unit":"percent","last_result":"2026-10-23","aliases":[],"url":"https://www.benchleader.com/benchmarks/aa_omniscience_non_hallucination"},{"id":"aa_briefcase","name":"AA-Briefcase v1.1","category":"agentic","category_label":"Agents & tools","measures":"Knowledge work delivered as real files — spreadsheets, decks, documents — graded head to head, so the score is an Elo rather than a percentage.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/aa-briefcase","models":66,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"Claude Sonnet 5.5 (max) at 1816","leader_model_url":"https://www.benchleader.com/models/claude-sonnet-5-5","leader_score":1815.77,"unit":"elo","last_result":"2026-10-23","aliases":[],"url":"https://www.benchleader.com/benchmarks/aa_briefcase"},{"id":"startrise_frontend","name":"Startrise Frontend","category":"coding","category_label":"Coding","measures":"Twelve single-shot frontend briefs, built in a real browser and scored on craft, engineering, brief adherence and originality. The blend is 25% deterministic gates, 45% a blind cross-lab judge panel and 30% a blind human review.","publisher":"Startrise","publisher_url":"https://www.startrise.io/benchmark/","models":29,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Claude Opus 5 at 82.3%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5","leader_score":82.3,"unit":"percent","last_result":"2026-10-05","aliases":[],"url":"https://www.benchleader.com/benchmarks/startrise_frontend"},{"id":"aa_frontiercode","name":"FrontierCode 1.1 (AA)","category":"coding","category_label":"Coding","measures":"Artificial Analysis' run of Cognition's FrontierCode. Epoch's run of the same benchmark is the one in the index.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/frontiercode","models":70,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Claude Opus 5.5 (medium) at 57.0%","leader_model_url":"https://www.benchleader.com/models/claude-opus-5-5","leader_score":57.05,"unit":"percent","last_result":"2026-10-23","aliases":[],"url":"https://www.benchleader.com/benchmarks/aa_frontiercode"},{"id":"aa_terminalbench_science","name":"Terminal-Bench-Science 0.1 (AA)","category":"agentic","category_label":"Agents & tools","measures":"Scientific computing tasks at a terminal: set up the environment, run the analysis, get the answer out. Run by Artificial Analysis.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/terminal-bench-science","models":82,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"GPT-6 Astra (max) at 63.3%","leader_model_url":"https://www.benchleader.com/models/gpt-6-astra","leader_score":63.33,"unit":"percent","last_result":"2026-10-23","aliases":[],"url":"https://www.benchleader.com/benchmarks/aa_terminalbench_science"},{"id":"aa_global_mmlu_lite","name":"Global-MMLU-Lite (AA)","category":"knowledge","category_label":"Knowledge","measures":"Knowledge questions across languages rather than only English. Run by Artificial Analysis; too few models so far to rank on.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/global-mmlu-lite","models":32,"in_index":false,"index_weight":null,"status":"Saturated","caveat":null,"leader":"Gemini 3.1 Pro at 93.2%","leader_model_url":"https://www.benchleader.com/models/gemini-3-1-pro","leader_score":93.17,"unit":"percent","last_result":"2026-03-17","aliases":[],"url":"https://www.benchleader.com/benchmarks/aa_global_mmlu_lite"},{"id":"aa_cyber_index","name":"AA Cyber Index","category":"composite","category_label":"Composite","measures":"Artificial Analysis' composite of its three security evaluations. Shown for orientation; the boards beneath it are what the index uses.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/aa-3d","models":21,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Kimi K3.1 Pro at 64.1","leader_model_url":"https://www.benchleader.com/models/kimi-k3-1-pro","leader_score":64.07,"unit":"index","last_result":"2026-10-23","aliases":[],"url":"https://www.benchleader.com/benchmarks/aa_cyber_index"},{"id":"aa_cwe_bench","name":"CWE-Bench (AA)","category":"agentic","category_label":"Agents & tools","measures":"Finding and fixing the vulnerability classes CWE catalogues, in real repositories. Run by Artificial Analysis.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/aa-3d","models":21,"in_index":true,"index_weight":0.5,"status":"Active","caveat":"Harness-dependent","leader":"DeepSeek V4.1 Flash 1017 (max) at 69.2%","leader_model_url":"https://www.benchleader.com/models/deepseek-v4-1-flash-1017","leader_score":69.17,"unit":"percent","last_result":"2026-10-23","aliases":[],"url":"https://www.benchleader.com/benchmarks/aa_cwe_bench"},{"id":"aa_deepsec_bench","name":"DeepSec-Bench (AA)","category":"agentic","category_label":"Agents & tools","measures":"Deeper security engineering tasks than a single patch: reason about a system and defend it. Run by Artificial Analysis.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/aa-3d","models":22,"in_index":true,"index_weight":0.5,"status":"Active","caveat":"Harness-dependent","leader":"GPT-6 Sol (max) at 46.3%","leader_model_url":"https://www.benchleader.com/models/gpt-6-sol","leader_score":46.32,"unit":"percent","last_result":"2026-10-23","aliases":[],"url":"https://www.benchleader.com/benchmarks/aa_deepsec_bench"},{"id":"aa_cybergym_e2e","name":"CyberGym end-to-end (AA)","category":"agentic","category_label":"Agents & tools","measures":"A full exploit-and-patch cycle in one run, scored end to end. Run by Artificial Analysis.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/aa-3d","models":21,"in_index":true,"index_weight":0.5,"status":"Near-saturated","caveat":"Harness-dependent","leader":"Kimi K3.1 Pro at 94.7%","leader_model_url":"https://www.benchleader.com/models/kimi-k3-1-pro","leader_score":94.66,"unit":"percent","last_result":"2026-10-23","aliases":[],"url":"https://www.benchleader.com/benchmarks/aa_cybergym_e2e"},{"id":"aa_3d","name":"AA-3D","category":"multimodal","category_label":"Multimodal","measures":"Interactive 3D experiences built from a prompt, graded head to head, so the score is an Elo rather than a percentage.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/aa-3d","models":22,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"Claude Sonnet 5.5 (max) at 1199","leader_model_url":"https://www.benchleader.com/models/claude-sonnet-5-5","leader_score":1198.83,"unit":"elo","last_result":"2026-10-23","aliases":[],"url":"https://www.benchleader.com/benchmarks/aa_3d"},{"id":"aa_openness_index","name":"AA Openness Index","category":"metadata","category_label":"Metadata","measures":"How open a model actually is: weights, licence, training details and reproducibility. A measure of openness, not capability, so it is never part of the quality index.","publisher":"Artificial Analysis","publisher_url":"https://artificialanalysis.ai/evaluations/artificial-analysis-openness-index","models":55,"in_index":false,"index_weight":null,"status":"Active","caveat":null,"leader":"K2 Think V2 (thinking) at 88.9","leader_model_url":"https://www.benchleader.com/models/k2-v2","leader_score":88.89,"unit":"index","last_result":"2026-10-17","aliases":[],"url":"https://www.benchleader.com/benchmarks/aa_openness_index"},{"id":"browsecomp_plus","name":"BrowseComp-Plus","category":"agentic","category_label":"Agents & tools","measures":"BrowseComp questions answered against a fixed 100k-document corpus instead of the live web, so the retriever and the model can be judged apart. Scores shown are on the BM25 baseline retriever, the one column nearly every submission runs. Model coverage stops in late 2025.","publisher":"BrowseComp-Plus","publisher_url":"https://github.com/texttron/BrowseComp-Plus","models":21,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"GPT-5 at 57.6%","leader_model_url":"https://www.benchleader.com/models/gpt-5","leader_score":57.59,"unit":"percent","last_result":"2025-12-31","aliases":["browsecomp plus","browsecomp+","browse comp plus","deep research"],"url":"https://www.benchleader.com/benchmarks/browsecomp_plus"},{"id":"kaggle_browsecomp","name":"BrowseComp","category":"agentic","category_label":"Agents & tools","measures":"Hard questions whose answers are buried deep on the open web: the model has to browse, not recall. Published by OpenAI.","publisher":"OpenAI","publisher_url":"https://openai.com/index/browsecomp/","models":39,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"Gemini 3.1 Pro at 31.2%","leader_model_url":"https://www.benchleader.com/models/gemini-3-1-pro","leader_score":31.19,"unit":"percent","last_result":"2026-03-25","aliases":["browsecomp","browse comp","web search","deep research"],"url":"https://www.benchleader.com/benchmarks/kaggle_browsecomp"},{"id":"kaggle_dsqa","name":"DeepSearchQA","category":"agentic","category_label":"Agents & tools","measures":"Research questions with many separate facts to gather; scored on answers that are fully correct, so missing one item fails the task. Published by Google.","publisher":"Google","publisher_url":"https://www.kaggle.com/benchmarks/google/dsqa","models":12,"in_index":false,"index_weight":null,"status":"Active","caveat":"Harness-dependent","leader":"Gemini Deep Research Agent at 66.1%","leader_model_url":"https://www.benchleader.com/models/gemini-deep-research-agent","leader_score":66.09,"unit":"percent","last_result":"2026-04-13","aliases":["dsqa","deepsearchqa","deep search qa","deep research"],"url":"https://www.benchleader.com/benchmarks/kaggle_dsqa"},{"id":"kaggle_facts_search","name":"FACTS Search","category":"agentic","category_label":"Agents & tools","measures":"Using web search as a tool and synthesising what comes back without inventing anything. Published by Google.","publisher":"Google","publisher_url":"https://www.kaggle.com/benchmarks/google/facts-search","models":22,"in_index":false,"index_weight":null,"status":"Near-saturated","caveat":"Harness-dependent","leader":"GPT-6 Astra at 89.9%","leader_model_url":"https://www.benchleader.com/models/gpt-6-astra","leader_score":89.92,"unit":"percent","last_result":"2026-10-09","aliases":["facts search","facts","web search"],"url":"https://www.benchleader.com/benchmarks/kaggle_facts_search"}]}