AnthropicClaude Sonnet 4.5
Claude Sonnet 4.5 is an Anthropic proprietary reasoning model, released 29 Sept 2025. Its best configuration (high reasoning effort) ranks #160 of 785 on the BenchLeader Index at 58.1 ±6.4, in the upper half. The ± is the point: 303 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 reasoning (65) and lowest in coding (55). At $6.00 per million tokens blended it is among the most expensive ranked models. Output speed of 38 tokens per second puts it in the slowest quarter, with a first token in 1.1 s. It has been measured at 5 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 22 Oct 2026.
- Blended price
- $6.00/M
- $3.00 in · $15.00 out
- Output speed
- 38 tok/s
- OpenRouter traffic, 7-day median; not yet measured by Artificial Analysis
- First answer
- 1.09 s
- Context
- 200k
- Full answer
- –
- Cost per run
- –
- Released
- 29 Sept 2025
- Overall index58
- Reasoning65
- Coding55
- Maths63
- Human preference64
Versions
Anthropic has shipped 10 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 Sonnet 5.5max | 28 Sept 2026 | 69.3 | #13 |
| Claude Sonnet 5high | 30 Jun 2026 | 60.2 | #114 |
| Claude Sonnet 4.6 | 17 Feb 2026 | 57.1 | #185 |
| Claude Sonnet 4.5highthis page | 29 Sept 2025 | 58.1 | #160 |
| Claude Sonnet 4thinking | 22 May 2025 | 53.2 | #294 |
| Claude 3.7 Sonnetthinking | 24 Feb 2025 | 51.0 | #353 |
| Claude 3.5 Sonnet | 22 Oct 2024 | 46.3 | #490 |
| Claude 3 Sonnet | 29 Feb 2024 | 38.0 | #721 |
| Claude 37 Sonnetthinking | – | – | – |
| Claude Sonnet 5.0max | – | – | – |
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 | 50.3 | #375 | 47 tok/s | 1.02 s | $0.0057 | Agents & tools 59 · Composite 51 · Instruction following 45 · Knowledge 57 · Long context 52 · Multimodal 48 · Reasoning 41 |
| low | – | – | 38 tok/s | 1.09 s | $0.0057 | |
| highbest | 58.1 | #160 | 38 tok/s | 1.09 s | $0.0057 | Coding 55 · Human preference 64 · Maths 63 · Reasoning 65 |
| thinking | 52.6 | #309 | 47 tok/s | 13 s | $0.0057 | Agents & tools 48 · Coding 52 · Composite 53 · Instruction following 50 · Knowledge 58 · Long context 61 · Maths 54 · Multimodal 57 · Reasoning 43 |
| not stated | 53.7 | #273 | 38 tok/s | 1.09 s | $0.0057 | Agents & tools 58 · Coding 53 · Human preference 64 · Instruction following 53 · Knowledge 45 · Maths 52 · Multimodal 59 · Reasoning 48 |
How its index has moved
8 Sept 2026 to 21 Oct 2026- high
- not stated
- thinking
- 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
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 | high | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| GPQA Diamondeffort not stated | – | – | – | 82.3%#113 | Epoch AI Benchmarking Hub | |
| Humanity's Last Exameffort not stated | – | – | – | 13.7%#27 | Scale AI / CAIS | |
| SimpleBencheffort not stated | – | – | – | 54.3%#40 | SimpleBench | |
| Arena Hard Prompts | – | 1487#55 | – | 1483#60 | Arena (LMArena) | |
| GPQA Diamond (AA)not in index | 72.7%#274 | – | 83.4%#159 | – | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 7.2%#371 | – | 17.8%#234 | – | Artificial Analysis | |
| GPQA Diamond (Vals)not in index | – | – | 81.6%#61 | – | Vals AI | |
| Kagi LLM Benchmarkeffort not stated | – | – | – | 57.9%#55 | Kagi LLM Benchmark | |
| GPQA Diamond (HELM)not in indexeffort not stated | – | – | – | 68.6%#11 | HELM Capabilities | |
| ARC-AGI-1 | – | – | 63.7%#155 | 25.5%#213 | ARC Prize | |
| ARC-AGI-2 | – | – | 13.6%#151 | 3.8%#189 | ARC Prize | |
| CritPt | 0.0%#315 | – | 1.1%#223 | – | Artificial Analysis | |
| SciPredictnot in indexeffort not stated | – | – | – | 22.6%#3 | Scale AI SEAL | |
| Chess Puzzlesnot in indexeffort not stated | – | – | – | 10.0%#119 | Epoch AI Benchmarking Hub | |
| Mystery Game Puzzlesnot in indexeffort not stated | – | – | – | 17.0%#75 | Epoch AI Benchmarking Hub | |
| LMCAnot in indexeffort not stated | – | – | – | 38.8%#111 | Epoch AI Benchmarking Hub | |
| DTBenchnot in indexeffort not stated | – | – | – | 83.2%#114 | Epoch AI Benchmarking Hub | |
| ForecastBenchnot in indexeffort not stated | – | – | – | 64.1%#35 | ForecastBench |
Coding
Agents & tools
| Benchmark | no reasoning | low | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| Terminal-Bencheffort not stated | – | – | – | 46.5%#39 | Terminal-Bench | |
| GDPvaleffort not stated | – | – | – | 42.5%#4 | OpenAI | |
| Cybencheffort not stated | – | – | – | 60.0%#3 | Cybench | |
| Remote Labor Indexeffort not stated | – | – | – | 2.1%#12 | Scale AI / CAIS | |
| Terminal-Bench Hard | 28.8%#117 | – | 35.6%#74 | – | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | 70.5%#155 | – | 78.1%#131 | – | Artificial Analysis | |
| MCP Atlas | – | – | 59.5%#27 | – | Scale AI SEAL | |
| BFCL Overalleffort not stated | – | – | – | 73.2%#2 | Berkeley Function Calling Leaderboard | |
| τ²-bencheffort not stated | – | – | – | 76.4%#7 | τ²-bench | |
| τ³-bench Bankingnot in indexeffort not stated | – | – | – | 25.3%#17 | τ²-bench | |
| τ³-Banking (AA) | – | – | 24.5%#85 | – | Artificial Analysis | |
| Poker Agentnot in indexeffort not stated | – | – | – | 1055.5%#7 | Vals AI | |
| Terminal-Bench 2.0 (Vals)not in index | – | – | 41.6%#29 | – | Vals AI | |
| DeepResearch Benchnot in index | – | 47.5%#23 | – | 52.6%#7 | DeepResearch Bench | |
| METR Time Horizonsnot in indexeffort not stated | – | – | – | 67.4%#12 | METR | |
| Vending-Bench 2not in indexeffort not stated | – | – | – | 3838.7#43 | Andon Labs | |
| AutomationBenchnot in index | – | – | 14.1%#146 | – | Artificial Analysis | |
| GDP.pdf (AA)not in index | – | – | 5.2%#163 | – | Artificial Analysis | |
| DeepSearchQAnot in index | – | – | 16.0%#11 | – |
Maths
| Benchmark | high | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| FrontierMath Tiers 1–3effort not stated | – | – | 23.9%#83 | Epoch AI Benchmarking Hub | |
| FrontierMath Tier 4effort not stated | – | – | 2.4%#68 | Epoch AI Benchmarking Hub | |
| OTIS Mock AIMEeffort not stated | – | – | 77.8%#122 | Epoch AI Benchmarking Hub | |
| MATH Level 5effort not stated | – | – | 97.7%#6 | Epoch AI Benchmarking Hub | |
| ProofBencheffort not stated | – | – | 19.0%#52 | Vals AI | |
| AIME (Vals) | – | 88.2%#30 | – | Vals AI | |
| MGSM | – | 94.3%#4 | – | Vals AI | |
| MathArena Apexpossible contamination | – | 1.6%#33 | – | MathArena | |
| Omni-MATH (HELM)effort not stated | – | – | 55.3%#14 | HELM Capabilities | |
| Arena Maths | 1455#68 | – | 1428#109 | Arena (LMArena) |
Knowledge
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| SimpleQA Verifiedeffort not stated | – | – | 30.7%#66 | Epoch AI Benchmarking Hub | |
| AA-Omniscience | -9.4#187 | -0.1#138 | – | Artificial Analysis | |
| MMLU-Pro | – | 87.4%#28 | – | Vals AI | |
| LegalBench | – | 84.1%#39 | – | Vals AI | |
| CorpFin | – | 62.0%#49 | 60.8%#63 | Vals AI | |
| TaxEval | – | 73.3%#49 | – | Vals AI | |
| MedQA | – | 94.7%#15 | – | Vals AI | |
| PRBench Financeeffort not stated | – | – | 43.8%#23 | Scale AI SEAL | |
| PRBench Legaleffort not stated | – | – | 40.8%#30 | Scale AI SEAL | |
| MultiNRCpossible contamination | – | 35.8%#20 | 28.1%#25 | Scale AI SEAL | |
| MMLU-Pro (HELM)not in indexeffort not stated | – | – | 86.9%#3 | HELM Capabilities | |
| CaseLaw v2not in index | – | 62.2%#19 | – | Vals AI | |
| MedCodenot in index | – | 44.1%#42 | 40.6%#61 | Vals AI | |
| MedScribenot in index | – | 84.1%#38 | 84.5%#33 | Vals AI | |
| EBR-benchnot in indexeffort not stated | – | – | 2.4%#26 | Epoch AI Benchmarking Hub | |
| AA-Omniscience: accuracynot in index | 28.4%#198 | 32.9%#168 | – | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | 47.3%#134 | 50.9%#112 | – | Artificial Analysis |
Instruction following
| Benchmark | no reasoning | high | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| IFBench | 42.6%#236 | – | 57.3%#139 | – | Artificial Analysis | |
| MultiChallenge | – | – | 55.3%#23 | – | Scale AI SEAL | |
| TutorBench | – | – | 49.0%#21 | 45.7%#24 | Scale AI SEAL | |
| IFEval (HELM)effort not stated | – | – | – | 85.0%#19 | HELM Capabilities | |
| WildBench (HELM)effort not stated | – | – | – | 85.4%#9 | HELM Capabilities | |
| Arena Instruction Followingnot in index | – | 1464#46 | – | 1462#49 | Arena (LMArena) |
Human preference
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| Arena Text | 1456#67 | 1455#72 | Arena (LMArena) | |
| Arena Creative Writingnot in index | 1450#38 | 1454#33 | Arena (LMArena) | |
| Arena Multi-turnnot in index | 1472#57 | 1477#46 | Arena (LMArena) | |
| Arena Longer Queriesnot in index | 1485#36 | 1482#40 | Arena (LMArena) |
Multimodal
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| MMMU-Pro | 65.2%#177 | 68.7%#162 | – | Artificial Analysis | |
| VISTA | – | 48.8%#14 | 45.0%#30 | Scale AI SEAL | |
| MMMU (validation)effort not stated | – | – | 77.8%#11 | MMMU | |
| MMMU-Pro (official)not in indexeffort not stated | – | – | 68.9%#10 | MMMU | |
| MMMU-Pro (Vals)not in index | – | 79.3%#47 | – | Vals AI | |
| MortgageTaxnot in index | – | 64.0%#48 | – | Vals AI | |
| SAGEnot in index | – | 36.1%#65 | 32.9%#71 | Vals AI | |
| VTBnot in index | – | 6.2%#11 | 5.6%#12 | Scale AI SEAL | |
| VPCTnot in indexeffort not stated | – | – | 39.8%#18 | Epoch AI Benchmarking Hub |
Long context
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| AA-LCR | 54.0%#305 | 72.3%#188 | – | Artificial Analysis | |
| Arena Documentnot in indexeffort not stated | – | – | 1449#33 | Arena (LMArena) | |
| MLCRnot in index | – | 10.6%#108 | – | Artificial Analysis |
Composite
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| Epoch Capabilities Indexnot in indexeffort not stated | – | – | 146.7#72 | Epoch AI Benchmarking Hub | |
| AA Intelligence Index v4.3.2 | 19.3#243 | 20.7#225 | – | Artificial Analysis | |
| HELM Capabilities meannot in indexeffort not stated | – | – | 76.2%#12 | HELM Capabilities |
Safety & honesty
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| FORTRESSnot in index | 12.8%#62 | 17.0%#48 | Scale AI SEAL | |
| MASKnot in indexpossible contamination | 96.1%#2 | 86.4%#15 | Scale AI SEAL |
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 |
|---|---|---|---|---|---|---|
| Anthropic | 43 tok/s | 0.86 s | $3.00 | $15.00 | 1M | – |
| Google Vertex | 39 tok/s | 2.37 s | $3.00 | $15.00 | 1M | – |
| Claude Platform on AWS | 38 tok/s | 1.07 s | $3.00 | $15.00 | 1M | – |
| Amazon Bedrock | 35 tok/s | 1.55 s | $3.00 | $15.00 | 1M | – |
| Azure | 34 tok/s | 3.95 s | $3.00 | $15.00 | 200k | – |
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 $0.300 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0057 | $0.0049 | 9.0 s |
| Summarize a 30-page report | 12,000 / 600 | $0.045 | $0.021 | 16.9 s |
| Code edit | 6,000 / 1,500 | $0.041 | $0.028 | 40.6 s |
| Agentic coding session | 60,000 / 4,000 | $0.240 | $0.118 | 1.8 min |
| Structured extraction | 2,000 / 200 | $0.0090 | $0.0050 | 6.4 s |
More from Anthropic
See also
Questions people ask
- What are Claude Sonnet 4.5's benchmark scores?
- As of 24 Oct 2026, Claude Sonnet 4.5 (high effort) scores 58.1 ±6.4 on the BenchLeader Index, strongest in reasoning (65) and weakest in coding (55). That rests on 6 independent benchmark results, and every score on this page links to its publisher.
- How does Claude Sonnet 4.5 rank?
- Claude Sonnet 4.5 (high effort) is #160 of 785 ranked configurations on 24 Oct 2026. 303 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 Sonnet 4.5 cost per 1M tokens?
- As of 24 Oct 2026, Claude Sonnet 4.5 costs $3.00 per 1M input tokens and $15.00 per 1M output tokens, $6.00 blended (three input tokens per output token).
- When was Claude Sonnet 4.5 released?
- Anthropic released Claude Sonnet 4.5 on 29 Sept 2025. BenchLeader has results for it at 5 reasoning-effort settings.
Data as of 24 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “Claude Sonnet 4.5: benchmarks, pricing, speed and rank”, data as of 24 Oct 2026, https://www.benchleader.com/models/claude-sonnet-4-5.