Muse Spark Open added (16 Oct 2026)
Muse Spark Open joins the board; Nemotron 3.5 Bolt 74B-A9B joins the board. 4 new models, 290 new benchmark results.
BenchLeader refreshed 1574 models from 24 of 24 data feeds (benchmark publishers, aggregators and price lists). 4 new models, 290 new or updated benchmark results across 24 data feeds.
New models
- Muse Spark Open (xhigh) (Meta) appeared with 5 benchmark results, entering the BenchLeader Index at #94. (details)
- Nemotron 3.5 Bolt 74B-A9B (NVIDIA) appeared with 4 benchmark results. (details)
- Nemotron 3.5 Super VL (NVIDIA) appeared with 4 benchmark results. (details)
- Glm Legal (max) (Zhipu AI) appeared with 1 benchmark result. (details)
New benchmark results
- Startrise Frontend: 29 new results, including Claude Opus 5.5, GPT-6 Astra, Claude Fable 5… (board)
- SREBench: 18 new results, including Muse Wonder Full, Muse Spark 1.3, GPT-6 Sol… (board)
- ProgramBench: 10 new results, including GPT-6.1 Sol, Muse Wonder Full, Muse Spark 1.3… (board)
- FrontierCode 1.1 (AA): 8 new results, including Claude Opus 5.5, Claude Opus 5.5, Claude Sonnet 5.5… (board)
- Vals Index: 8 new results, including Muse Wonder Full, Muse Wonder, Kimi K3… (board)
- Legal Research Bench: 8 new results, including Muse Wonder Full, Muse Wonder, Kimi K3… (board)
- Terminal-Bench Science: 7 new results, including GPT-6.1 Sol, Muse Wonder Full, Muse Wonder… (board)
- Vibe Code Bench v1.1: 7 new results, including Muse Wonder Full, Muse Wonder, Kimi K3… (board)
- IOI: 6 new results, including Muse Wonder Full, Muse Wonder, Step 5 Preview… (board)
- BioMysteryBench: 6 new results, including Muse Wonder Full, Muse Wonder, Step 5 Preview… (board)
- Code Migration: 6 new results, including Muse Wonder Full, Muse Wonder, Step 5 Preview… (board)
- Finance Agent v2: 6 new results, including Muse Wonder Full, Muse Wonder, Kimi K3… (board)
Score revisions
- Gemini 4 Argon (high) on Tax Agent Bench: 76.2% → 44.6%. (details)
- GPT-6 Astra (max) on Tax Agent Bench: 63.3% → 20.7%. (details)
- GPT-6.1 Sol (max) on Tax Agent Bench: 62.3% → 21.8%. (details)
- Claude Fable 5 on Tax Agent Bench: 69.8% → 38.3%. (details)
- GPT-5.6 Sol (max) on Tax Agent Bench: 68% → 31.1%. (details)
- Claude Sonnet 5.5 (max) on Arena Agent: 13 → 12.5. (details)
- Muse Spark 1.3 (max) on Tax Agent Bench: 72.4% → 42.4%. (details)
- GPT-5.5 (xhigh) on Tax Agent Bench: 60.5% → 21.8%. (details)
- GPT-6 Sol (max) on Arena Agent: 10.5 → 11.2. (details)
- GPT-6 Sol (max) on Tax Agent Bench: 53.1% → 15%. (details)
- Claude Opus 5 on Tax Agent Bench: 75.1% → 45.6%. (details)
- Claude Fable 5.1 on Tax Agent Bench: 77.6% → 49.2%. (details)
- MiMo-V2.6-Pro on Tax Agent Bench: 64.9% → 32.4%. (details)
- Kimi K3 (max) on BioMysteryBench: 71.5% → 72.6%. (details)
- Kimi K3 (max) on Harvey's Legal Agent Benchmark: 10.8% → 12.9%. (details)
- Kimi K3 (max) on Tax Agent Bench: 68.7% → 33.4%. (details)
- Kimi K3 (max) on Terminal-Bench 4.0 (Vals): 19.7% → 17.2%. (details)
- Kimi K3 (max) on Terminal-Bench Science: 2.9% → 1.4%. (details)
- Muse Spark 1.3 (xhigh) on Tax Agent Bench: 71.9% → 36%. (details)
- Qwen3.8 Max (max) on Tax Agent Bench: 66% → 32.1%. (details)
- …and 73 more.
Speed changes
- GPT-6 Astra time to first answer changed from 94.34 s to 58.46 s. (details)
- Claude Fable 5.1 time to first answer changed from 20.37 s to 9.72 s. (details)
- Claude Fable 5.1 response time changed from 29.42 s to 17.39 s. (details)
- Claude Opus 5 time to first answer changed from 28.46 s to 17.79 s. (details)
- Claude Opus 5 response time changed from 108.92 s to 67.80 s. (details)
- GPT-6.1 Sol time to first answer changed from 7.59 s to 4.78 s. (details)
- Muse Wonder output speed changed from 115 tok/s to 161 tok/s. (details)
- GPT-5.6 Sol time to first answer changed from 60.43 s to 38.65 s. (details)
- GPT-5.6 Sol response time changed from 34.95 s to 14.35 s. (details)
- GPT-5.5 time to first answer changed from 70.60 s to 37.05 s. (details)
- GPT-5.5 response time changed from 75.52 s to 42.12 s. (details)
- GPT-6 Sol time to first token changed from 6.55 s to 2.66 s. (details)
- GPT-5.5 Pro output speed changed from 11 tok/s to 22 tok/s. (details)
- Kimi K3 time to first token changed from 0.86 s to 1.45 s. (details)
- Claude Opus 5.5 time to first answer changed from 12.35 s to 7.70 s. (details)
- …and 115 more.
Today's top five
- Claude Opus 5.5 (max) — 72.6
- Gemini 4 Argon (high) — 70.7
- Muse Wonder Full (max) — 70.6
- Claude Fable 5.1 (max) — 70.5
- GPT-6 Astra (max) — 70.4
More: New AI model releases this week, and the full LLM leaderboard 2026.
This digest is generated automatically from the day's data changes. Every line links to the page where you can check the numbers and their source.