AnthropicClaude Opus 4
Claude Opus 4 is an Anthropic proprietary reasoning model, released 22 May 2025. Its best configuration (thinking reasoning effort) ranks #236 of 785 on the BenchLeader Index at 55.0 ±3.9, in the upper half. The ± is the point: 273 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in maths (65) and lowest in reasoning (48). At $30.00 per million tokens blended it is among the most expensive ranked models. It has been measured at 3 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 22 Oct 2026.
- Blended price
- $30.00/M
- $15.00 in · $75.00 out
- Output speed
- –
- First answer
- –
- Context
- 200k
- Full answer
- –
- Cost per run
- –
- Released
- 22 May 2025
- Overall index55
- Reasoning48
- Coding54
- Agents & tools61
- Maths65
- Knowledge49
- Instruction following54
- Human preference61
- Multimodal55
- Long context60
- Composite53
Versions
Anthropic has shipped 9 models under this name. Each is ranked on its own results; a newer version often has fewer results so far, which holds its index nearer the average until more arrive.
| Model | Released | Index | Rank |
|---|---|---|---|
| Claude Opus 5.5max | 22 Sept 2026 | 72.6 | #1 |
| Claude Opus 5max | 24 Jul 2026 | 68.8 | #15 |
| Claude Opus 4.8max | 28 May 2026 | 63.3 | #67 |
| Claude Opus 4.7 | 16 Apr 2026 | 63.3 | #66 |
| Claude Opus 4.6high | 5 Feb 2026 | 62.7 | #78 |
| Claude Opus 4.5thinking | 24 Nov 2025 | 59.9 | #119 |
| Claude Opus 4.1thinking | 5 Aug 2025 | 56.4 | #204 |
| Claude Opus 4thinkingthis page | 22 May 2025 | 55.0 | #236 |
| Claude 3 Opus | 4 Mar 2024 | 39.2 | #691 |
Reasoning-effort configurations
The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Not stated” collects results from publishers that did not say which setting they used; for a reasoning model that is usually its thinking mode, but we do not assume it. Pick a setting here or at the top of the page to see its price, speed and category scores.
| Effort | Index | Rank | Speed | First answer | Chat reply cost | Categories |
|---|---|---|---|---|---|---|
| no reasoning | – | – | – | – | $0.029 | Composite 48 · Instruction following 46 |
| thinkingbest | 55.0 | #236 | – | – | $0.029 | Agents & tools 61 · Coding 54 · Composite 53 · Human preference 61 · Instruction following 54 · Knowledge 49 · Long context 60 · Maths 65 · Multimodal 55 · Reasoning 48 |
| not stated | 52.6 | #308 | – | – | $0.029 | Agents & tools 56 · Coding 53 · Human preference 59 · Instruction following 58 · Knowledge 55 · Long context 37 · Maths 57 · Multimodal 53 · Reasoning 47 |
How its index has moved
8 Sept 2026 to 21 Oct 2026- thinking
- not stated
The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.
Benchmark results
See all LLM benchmarks. One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.
Reasoning
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| GPQA Diamondeffort not stated | – | – | 76.3%#142 | Epoch AI Benchmarking Hub | |
| Humanity's Last Exameffort not stated | – | – | 10.7%#30 | Scale AI / CAIS | |
| SimpleBencheffort not stated | – | – | 58.8%#32 | SimpleBench | |
| Arena Hard Prompts | – | 1457#102 | 1436#128 | Arena (LMArena) | |
| GPQA Diamond (AA)not in index | 70.1%#295 | 79.6%#197 | – | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 6.2%#405 | 12.3%#280 | – | Artificial Analysis | |
| GPQA Diamond (Vals)not in indexeffort not stated | – | – | 71.7%#86 | Vals AI | |
| Kagi LLM Benchmark | – | 74.3%#15 | 59.6%#52 | Kagi LLM Benchmark | |
| GPQA Diamond (HELM)not in index | – | 70.9%#9 | 66.6%#15 | HELM Capabilities | |
| ARC-AGI-1 | – | 35.7%#198 | 22.5%#216 | ARC Prize | |
| ARC-AGI-2 | – | 8.6%#164 | 1.3%#218 | ARC Prize | |
| LMCAnot in indexeffort not stated | – | – | 37.4%#118 | Epoch AI Benchmarking Hub | |
| DTBenchnot in indexeffort not stated | – | – | 81.6%#124 | Epoch AI Benchmarking Hub |
Coding
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| SWE-bench Verified (Epoch)effort not stated | – | 70.7%#25 | Epoch AI Benchmarking Hub | |
| WeirdMLeffort not stated | – | 43.7%#86 | WeirdML | |
| GSO-Bencheffort not stated | – | 6.9%#23 | GSO-Bench | |
| Arena Coding | 1498#82 | 1465#122 | Arena (LMArena) | |
| LiveCodeBench | 70.2%#88 | 62.6%#98 | Vals AI | |
| Aider Polygloteffort not stated | – | 72.0%#7 | Aider polyglot leaderboard | |
| SWE-bench Verified (bash only)effort not stated | – | 67.6%#15 | SWE-bench | |
| SWE-bench Verified (any scaffold)not in indexeffort not stated | – | 73.2%#9 | SWE-bench | |
| AlgoTunenot in indexeffort not stated | – | 1.3#17 | AlgoTune |
Agents & tools
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| Cybencheffort not stated | – | 38.0%#7 | Cybench | |
| Terminal-Bench Hard | 31.1%#106 | – | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | 73.4%#145 | – | Artificial Analysis | |
| DeepResearch Benchnot in indexeffort not stated | – | 46.8%#26 | DeepResearch Bench | |
| METR Time Horizonsnot in indexeffort not stated | – | 63.9%#16 | METR | |
| BrowseComp-Plusnot in indexeffort not stated | – | 15.5%#12 | BrowseComp-Plus | |
| BrowseCompnot in indexeffort not stated | – | 1.0%#16 | OpenAI |
Maths
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| OTIS Mock AIMEeffort not stated | – | 64.4%#155 | Epoch AI Benchmarking Hub | |
| MATH Level 5effort not stated | – | 85.0%#27 | Epoch AI Benchmarking Hub | |
| AIME (Vals)effort not stated | – | 41.3%#66 | Vals AI | |
| MGSMeffort not stated | – | 93.8%#8 | Vals AI | |
| Omni-MATH (HELM) | 61.6%#8 | 51.1%#18 | HELM Capabilities | |
| MATH 500not in indexeffort not stated | – | 90.4%#24 | Vals AI | |
| Arena Maths | 1421#120 | 1403#146 | Arena (LMArena) |
Knowledge
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| MMLU-Proeffort not stated | – | 86.2%#46 | Vals AI | |
| LegalBencheffort not stated | – | 83.1%#59 | Vals AI | |
| TaxEvaleffort not stated | – | 71.9%#68 | Vals AI | |
| MedQAeffort not stated | – | 92.9%#26 | Vals AI | |
| MultiNRC | 33.9%#22 | 29.0%#24 | Scale AI SEAL | |
| MMLU-Pro (HELM)not in index | 87.5%#2 | 85.9%#6 | HELM Capabilities |
Instruction following
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| IFBench | 43.3%#225 | 53.7%#155 | – | Artificial Analysis | |
| MultiChallenge | – | 58.6%#17 | – | Scale AI SEAL | |
| TutorBench | – | 49.7%#19 | 45.5%#25 | Scale AI SEAL | |
| IFEval (HELM) | – | 84.9%#21 | 91.8%#7 | HELM Capabilities | |
| WildBench (HELM) | – | 85.2%#12 | 83.3%#18 | HELM Capabilities | |
| Arena Instruction Followingnot in index | – | 1444#72 | 1415#116 | Arena (LMArena) |
Human preference
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| Arena Text | 1426#115 | 1414#135 | Arena (LMArena) | |
| Arena Creative Writingnot in index | 1433#69 | 1417#85 | Arena (LMArena) | |
| Arena Multi-turnnot in index | 1439#104 | 1431#113 | Arena (LMArena) | |
| Arena Longer Queriesnot in index | 1466#66 | 1447#91 | Arena (LMArena) | |
| Lech Mazur Creative Writingnot in indexeffort not stated | – | 83.6%#7 | Lech Mazur |
Multimodal
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| Arena Vision | 1192#96 | 1176#104 | Arena (LMArena) | |
| VISTA | 47.0%#18 | 43.5%#32 | Scale AI SEAL | |
| MMMU (validation)effort not stated | – | 76.5%#13 | MMMU | |
| MMMU-Pro (Vals)not in indexeffort not stated | – | 73.3%#58 | Vals AI | |
| MortgageTaxnot in indexeffort not stated | – | 58.6%#66 | Vals AI | |
| GeoBenchnot in indexeffort not stated | – | 49.0%#23 | GeoBench | |
| VPCTnot in indexeffort not stated | – | 38.0%#21 | Epoch AI Benchmarking Hub |
Long context
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| Fiction.LiveBench 120keffort not stated | – | 37.5%#26 | Fiction.live | |
| AA-LCR | 69.3%#226 | – | Artificial Analysis |
Composite
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| Epoch Capabilities Indexnot in indexeffort not stated | – | – | 142.5#99 | Epoch AI Benchmarking Hub | |
| AA Intelligence Index v4.3.2 | 16.6#273 | 20.6#226 | – | Artificial Analysis | |
| HELM Capabilities meannot in index | – | 78.0%#8 | 75.7%#13 | HELM Capabilities |
Safety & honesty
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| FORTRESSnot in index | 24.8%#32 | 27.6%#27 | Scale AI SEAL | |
| MASKnot in indexpossible contamination | 87.9%#10 | 80.3%#24 | Scale AI SEAL |
What a task costs
Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache at $1.50 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.029 | $0.025 | – |
| Summarize a 30-page report | 12,000 / 600 | $0.225 | $0.103 | – |
| Code edit | 6,000 / 1,500 | $0.203 | $0.142 | – |
| Agentic coding session | 60,000 / 4,000 | $1.20 | $0.593 | – |
| Structured extraction | 2,000 / 200 | $0.045 | $0.025 | – |
More from Anthropic
See also
Questions people ask
- What are Claude Opus 4's benchmark scores?
- As of 24 Oct 2026, Claude Opus 4 (thinking effort) scores 55.0 ±3.9 on the BenchLeader Index, strongest in maths (65) and weakest in reasoning (48). That rests on 20 independent benchmark results, and every score on this page links to its publisher.
- How does Claude Opus 4 rank?
- Claude Opus 4 (thinking effort) is #236 of 785 ranked configurations on 24 Oct 2026. 273 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Opus 5.5 (max effort) at 72.6.
- How much does Claude Opus 4 cost per 1M tokens?
- As of 24 Oct 2026, Claude Opus 4 costs $15.00 per 1M input tokens and $75.00 per 1M output tokens, $30.00 blended (three input tokens per output token).
- When was Claude Opus 4 released?
- Anthropic released Claude Opus 4 on 22 May 2025. BenchLeader has results for it at 3 reasoning-effort settings.
Data as of 24 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “Claude Opus 4: benchmarks, pricing, speed and rank”, data as of 24 Oct 2026, https://www.benchleader.com/models/claude-opus-4.