AnthropicReasoning modelClaude Opus 4.5
Claude Opus 4.5 is an Anthropic proprietary reasoning model, released 24 Nov 2025. Its best configuration (thinking reasoning effort) ranks #106 of 756 on the BenchLeader Index at 60.2 ±2.6, in the upper half. The ± is the point: 184 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 agents & tools (75) and lowest in reasoning (55). At $10.00 per million tokens blended it is among the most expensive ranked models. Output speed of 49 tokens per second puts it slower than most, with a first answer in 16.8 s. It has been measured at 6 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 29 Sept 2026.
- Blended price
- $10.00/M
- $5.00 in · $25.00 out
- Output speed
- 49 tok/s
- measured by Artificial Analysis
- First answer
- 17 s
- first token 1.59 s
- Context
- 200k
- Full answer
- 27 s
- median, reasoning included
- Cost per run
- –
- Released
- 24 Nov 2025
- Overall index60
- Reasoning55
- Coding61
- Agents & tools75
- Maths64
- Knowledge60
- Instruction following56
- Multimodal57
- Long context64
- Composite63
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 5.5high | 22 Sept 2026 | 70.1 | #2 |
| Claude Opus 5high | 24 Jul 2026 | 68.4 | #14 |
| Claude Opus 4.8max | 28 May 2026 | 62.4 | #82 |
| Claude Opus 4.7 | 16 Apr 2026 | 63.8 | #59 |
| Claude Opus 4.6 | 17 Dec 2025 | 63.1 | #68 |
| Claude Opus 4.5thinkingthis page | 24 Nov 2025 | 60.2 | #106 |
| Claude Opus 4.1thinking | 5 Aug 2025 | 56.7 | #181 |
| Claude Opus 4thinking | 22 May 2025 | 55.3 | #214 |
| Claude 3 Opus | 4 Mar 2024 | 39.3 | #664 |
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 | 53.7 | #263 | 48 tok/s | 1.20 s | $0.0095 | Agents & tools 69 · Composite 57 · Instruction following 45 · Knowledge 60 · Long context 61 · Multimodal 54 · Reasoning 39 |
| low | – | – | 37 tok/s | 1.59 s | $0.0095 | |
| medium | – | – | 37 tok/s | 1.59 s | $0.0095 | Coding 65 |
| high | 57.2 | #174 | 37 tok/s | 1.59 s | $0.0095 | Agents & tools 54 · Coding 62 · Composite 43 · Human preference 67 · Maths 66 · Reasoning 67 |
| thinkingbest | 60.2 | #106 | 49 tok/s | 17 s | $0.0095 | Agents & tools 75 · Coding 61 · Composite 63 · Instruction following 56 · Knowledge 60 · Long context 64 · Maths 64 · Multimodal 57 · Reasoning 55 |
| not stated | 59.7 | #117 | 37 tok/s | 1.59 s | $0.0095 | Agents & tools 70 · Coding 57 · Human preference 66 · Knowledge 56 · Maths 53 · Multimodal 60 · Reasoning 61 |
How its index has moved
8 Sept 2026 to 5 Oct 2026- thinking
- not stated
- high
- no reasoning
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
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
Coding
| Benchmark | no reasoning | low | medium | high | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| SWE-bench Verified (Epoch) | – | – | – | – | – | 76.7%#9 | Epoch AI Benchmarking Hub | |
| WeirdML | – | – | – | – | – | 63.7%#41 | WeirdML | |
| GSO-Bench | – | – | – | – | – | 26.5%#14 | GSO-Bench | |
| LMArena Coding | – | – | – | 1530#20 | – | 1523#30 | LMArena | |
| LMArena WebDev | – | – | – | 1493#53 | – | 1469#60 | LMArena | |
| LiveBench Codingnot in index | – | – | – | 79.7%#22 | – | – | LiveBench | |
| LiveCodeBench | – | – | – | – | 83.7%#45 | 75.0%#78 | Vals AI | |
| SWE-bench (Vals)not in index | – | – | – | – | 76.4%#35 | – | Vals AI | |
| SWE-Bench Pro | – | – | – | – | – | 45.9%#6 | Scale AI SEAL | |
| SWE-bench Verified (bash only) | – | – | 74.4%#5 | 76.8%#1 | – | – | SWE-bench | |
| SWE-bench Verified (any scaffold)not in index | – | – | – | 76.8%#1 | – | – | SWE-bench | |
| Vibe Code Bench v1.1not in index | – | – | – | – | 20.6%#69 | – | Vals AI | |
| SWE-Bench Pro (private)not in index | – | – | – | – | – | 45.9%#6 | Scale AI SEAL | |
| ALE-Benchnot in index | – | – | – | – | – | 1025.4#48 | Epoch AI Benchmarking Hub | |
| AlgoTunenot in index | – | – | – | – | – | 1.8#5 | AlgoTune |
Agents & tools
| Benchmark | no reasoning | low | medium | high | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| Terminal-Bench | – | – | – | – | – | 63.1%#12 | Terminal-Bench | |
| GDPval | – | – | – | – | – | 45.5%#2 | OpenAI | |
| Cybench | – | – | – | – | – | 82.0%#2 | Cybench | |
| Remote Labor Index | – | – | – | – | – | 3.8%#9 | Scale AI / CAIS | |
| LiveBench Agentic Codingnot in index | – | – | – | 39.7%#58 | – | – | LiveBench | |
| Terminal-Bench Hard | 40.9%#53 | – | – | – | 47.0%#27 | – | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | 86.3%#88 | – | – | – | 89.5%#68 | – | Artificial Analysis | |
| MCP Atlas | – | – | – | 69.8%#21 | – | – | Scale AI SEAL | |
| BFCL Overall | – | – | – | – | – | 77.5%#1 | Berkeley Function Calling Leaderboard | |
| τ²-bench | – | – | – | 85.3%#2 | – | – | τ²-bench | |
| Poker Agentnot in index | – | – | – | – | – | 1033.4%#10 | Vals AI | |
| Terminal-Bench 2.0 (Vals)not in index | – | – | – | – | 53.9%#16 | 58.4%#10 | Vals AI | |
| BALROGnot in index | – | – | – | – | – | 43.5%#10 | BALROG | |
| DeepResearch Benchnot in index | – | 53.7%#5 | – | 54.8%#3 | – | – | DeepResearch Bench | |
| METR Time Horizonsnot in index | – | – | – | – | – | 75.0%#5 | METR | |
| Vending-Bench 2not in index | – | – | – | – | – | 4967.1#32 | Andon Labs | |
| DeepSearchQAnot in index | – | – | – | – | 24.0%#10 | – |
Maths
| Benchmark | no reasoning | low | medium | high | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| FrontierMath Tiers 1–3 | – | – | – | – | – | 34.4%#61 | Epoch AI Benchmarking Hub | |
| FrontierMath Tier 4 | – | – | – | – | – | 4.9%#59 | Epoch AI Benchmarking Hub | |
| OTIS Mock AIME | – | – | – | – | – | 86.1%#86 | Epoch AI Benchmarking Hub | |
| ProofBench | – | – | – | – | – | 36.0%#37 | Vals AI | |
| LiveBench Mathematicsnot in index | – | – | – | 90.4%#32 | – | – | LiveBench | |
| AIME (Vals) | – | – | – | – | 95.4%#12 | 76.9%#50 | Vals AI | |
| MGSM | – | – | – | – | 95.2%#1 | 94.8%#2 | Vals AI | |
| LMArena Maths | – | – | – | 1472#45 | – | 1463#57 | LMArena |
Knowledge
| Benchmark | no reasoning | low | medium | high | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| SimpleQA Verified | – | – | – | – | – | 45.7%#36 | Epoch AI Benchmarking Hub | |
| LiveBench Data Analysisnot in index | – | – | – | 74.4%#36 | – | – | LiveBench | |
| AA-Omniscience | -4.1#143 | – | – | – | 14#81 | – | Artificial Analysis | |
| MMLU-Pro | – | – | – | – | 87.3%#31 | 85.6%#55 | Vals AI | |
| LegalBench | – | – | – | – | 84.6%#29 | 82.8%#61 | Vals AI | |
| CorpFin | – | – | – | – | 65.1%#34 | 61.3%#54 | Vals AI | |
| TaxEval | – | – | – | – | 74.9%#25 | 74.3%#37 | Vals AI | |
| MedQA | – | – | – | – | 95.9%#10 | 93.2%#24 | Vals AI | |
| PRBench Finance | – | – | – | – | 46.2%#17 | – | Scale AI SEAL | |
| PRBench Legal | – | – | – | – | 44.2%#19 | – | Scale AI SEAL | |
| MultiNRC | – | – | – | – | 48.6%#12 | 41.2%#18 | Scale AI SEAL | |
| CaseLaw v2not in index | – | – | – | – | 62.6%#18 | – | Vals AI | |
| MedCodenot in index | – | – | – | – | 49.2%#22 | 45.2%#36 | Vals AI | |
| MedScribenot in index | – | – | – | – | 85.3%#26 | 83.3%#43 | Vals AI | |
| EBR-benchnot in index | – | – | – | – | – | 14.3%#15 | Epoch AI Benchmarking Hub | |
| AA-Omniscience: accuracynot in index | 40.9%#111 | – | – | – | 46.6%#78 | – | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | 23.8%#239 | – | – | – | 39.0%#152 | – | Artificial Analysis |
Instruction following
| Benchmark | no reasoning | low | medium | high | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| LiveBench Languagenot in index | – | – | – | 81.3%#26 | – | – | LiveBench | |
| LiveBench Instruction Followingnot in index | – | – | – | 62.5%#51 | – | – | LiveBench | |
| IFBench | 43.0%#228 | – | – | – | 58.0%#131 | – | Artificial Analysis | |
| MultiChallenge | – | – | – | – | 59.0%#12 | – | Scale AI SEAL | |
| LMArena Instruction Followingnot in index | – | – | – | 1483#19 | – | 1475#27 | LMArena |
Human preference
| Benchmark | no reasoning | low | medium | high | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| LMArena Text | – | – | – | 1474#39 | – | 1470#47 | LMArena | |
| LMArena Creative Writingnot in index | – | – | – | 1470#16 | – | 1462#25 | LMArena | |
| LMArena Multi-turnnot in index | – | – | – | 1486#29 | – | 1484#30 | LMArena | |
| LMArena Longer Queriesnot in index | – | – | – | 1495#20 | – | 1493#23 | LMArena |
Multimodal
| Benchmark | no reasoning | low | medium | high | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| MMMU-Pro | 71.2%#132 | – | – | – | 74.0%#110 | – | Artificial Analysis | |
| VISTA | – | – | – | – | 46.4%#21 | 45.3%#27 | Scale AI SEAL | |
| MMMU (validation) | – | – | – | – | – | 80.7%#5 | MMMU | |
| MMMU-Pro (official)not in index | – | – | – | – | – | 73.9%#8 | MMMU | |
| MMMU-Pro (Vals)not in index | – | – | – | – | 83.0%#36 | 81.1%#42 | Vals AI | |
| MortgageTaxnot in index | – | – | – | – | 67.7%#23 | 68.7%#12 | Vals AI | |
| SAGEnot in index | – | – | – | – | 52.1%#8 | 45.0%#41 | Vals AI | |
| GeoBenchnot in index | – | – | – | – | – | 75.0%#9 | GeoBench | |
| VPCTnot in index | – | – | – | – | – | 40.0%#16 | Epoch AI Benchmarking Hub |
Long context
| Benchmark | no reasoning | low | medium | high | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| AA-LCR | 70.7%#193 | – | – | – | 77.3%#120 | – | Artificial Analysis | |
| LMArena Documentnot in index | – | – | – | – | – | 1465#21 | LMArena |
Composite
| Benchmark | no reasoning | low | medium | high | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| Epoch Capabilities Indexnot in index | – | – | – | – | – | 150.1#48 | Epoch AI Benchmarking Hub | |
| LiveBench | – | – | – | 72.6%#48 | – | – | LiveBench | |
| AA Intelligence Index | 23.7#168 | – | – | – | 29.1#115 | – | Artificial Analysis |
Safety & honesty
| Benchmark | no reasoning | low | medium | high | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| FORTRESSnot in index | – | – | – | – | 9.6%#64 | 13.6%#56 | Scale AI SEAL | |
| MASKnot in index | – | – | – | – | 92.5%#5 | 87.1%#12 | Scale AI SEAL |
Where it wins
Benchmarks where this configuration ranks in the top five of every configuration measured.
Where to run it
Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.
| Provider | Speed | First token | Input $/M | Output $/M | Context | Quantisation |
|---|---|---|---|---|---|---|
| Google Vertex | 57 tok/s | 7.07 s | $5.00 | $25.00 | 200k | – |
| Anthropic | 50 tok/s | 1.09 s | $5.00 | $25.00 | 200k | – |
| Amazon Bedrock | 48 tok/s | 2.12 s | $5.50 | $27.50 | 200k | – |
| Amazon Bedrock | 44 tok/s | 1.56 s | $5.00 | $25.00 | 200k | – |
Price history
Listed price per 1M tokens over time, as recorded by OpenRouter for the provider with the longest history.
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 $0.500 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0095 | $0.0081 | 23.0 s |
| Summarise a 30-page report | 12,000 / 600 | $0.075 | $0.035 | 29.2 s |
| Code edit | 6,000 / 1,500 | $0.068 | $0.047 | 47.7 s |
| Agentic coding session | 60,000 / 4,000 | $0.400 | $0.198 | 1.7 min |
| Structured extraction | 2,000 / 200 | $0.015 | $0.0083 | 21.0 s |
See also
Data as of 5 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “Claude Opus 4.5: benchmarks, pricing, speed and rank”, https://www.benchleader.com/models/claude-opus-4-5, data as of 5 Oct 2026.