AnthropicClaude Sonnet 5.5
Claude Sonnet 5.5 is an Anthropic proprietary reasoning model, released 28 Sept 2026. Its best configuration (max reasoning effort) ranks #21 of 760 on the BenchLeader Index at 67.6 ±5.6, in the top 25. The ± is the point: 87 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 (81) and lowest in knowledge (65). At $4.00 per million tokens blended it is among the most expensive ranked models. Output speed of 141 tokens per second puts it in the fastest quarter, with a first answer in 487.1 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 8 Oct 2026.
- Blended price
- $4.00/M
- $2.00 in · $10.00 out
- Output speed
- 141 tok/s
- measured by Artificial Analysis
- First answer
- 487 s
- first token 2.69 s
- Context
- 1M
- Full answer
- 491 s
- median, reasoning included
- Cost per run
- $5.46
- one full Intelligence Index run
- Released
- 28 Sept 2026
- Overall index68
- Reasoning81
- Coding70
- Agents & tools70
- Maths73
- Knowledge65
- Long context67
- Composite73
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 Sonnet 5.5maxthis page | 28 Sept 2026 | 67.6 | #21 |
| Claude Sonnet 5high | 30 Jun 2026 | 60.5 | #102 |
| Claude Sonnet 4.6 | 17 Feb 2026 | 58.6 | #136 |
| Claude Sonnet 4.5high | 29 Sept 2025 | 58.3 | #144 |
| Claude Sonnet 4thinking | 22 May 2025 | 53.3 | #275 |
| Claude 3.7 Sonnetthinking | 24 Feb 2025 | 51.1 | #336 |
| Claude 3.5 Sonnet | 22 Oct 2024 | 46.4 | #468 |
| Claude 3 Sonnet | 29 Feb 2024 | 38.2 | #692 |
| Claude 37 Sonnetthinking | – | – | – |
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 |
|---|---|---|---|---|---|---|
| low | 58.4 | #143 | 100 tok/s | 0.88 s | $0.0038 | Coding 54 · Composite 70 · Knowledge 70 · Long context 63 · Reasoning 60 |
| medium | 59.1 | #124 | 99 tok/s | 1.87 s | $0.0038 | Agents & tools 44 · Coding 59 · Composite 76 · Knowledge 71 · Long context 63 · Reasoning 69 |
| high | 66.1 | #32 | 101 tok/s | 12 s | $0.0038 | Coding 70 · Composite 83 · Knowledge 71 · Long context 64 · Reasoning 82 |
| xhigh | 67.1 | #26 | 104 tok/s | 30 s | $0.0038 | Coding 72 · Composite 73 · Human preference 67 · Knowledge 72 · Long context 65 · Maths 70 · Multimodal 63 · Reasoning 85 |
| maxbest | 67.6 | #21 | 141 tok/s | 487 s | $0.0038 | Agents & tools 70 · Coding 70 · Composite 73 · Knowledge 65 · Long context 67 · Maths 73 · Reasoning 81 |
| not stated | – | – | 97 tok/s | 2.69 s | $0.0038 | Agents & tools 69 · Coding 65 |
What more thinking costs
Turning the effort up buys index points and multiplies the bill. Here max costs 16× low for +9.2 on the index — 2.3 points per doubling of spend. Compare that with other models.
How its index has moved
29 Sept 2026 to 9 Oct 2026- max
- xhigh
- high
- medium
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
| Benchmark | low | medium | high | xhigh | max | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| GPQA Diamond | – | – | – | – | 95.6%#2 | – | Epoch AI Benchmarking Hub | |
| LMArena Hard Prompts | – | – | – | 1507#19 | – | – | LMArena | |
| LiveBench Reasoningnot in index | – | – | – | 86.8%#29 | 91.6%#6 | – | LiveBench | |
| Humanity's Last Exam (AA)not in index | 36.2%#105 | 39.8%#82 | 45.8%#44 | 50.0%#23 | 55.0%#9 | – | Artificial Analysis | |
| CritPt | 11.4%#103 | 16.9%#77 | 24.6%#44 | 31.1%#9 | 31.4%#7 | – | Artificial Analysis | |
| MysteryMechanismnot in indexeffort not stated | – | – | – | – | – | 49.1%#3 | Vals AI | |
| Mystery Game Puzzlesnot in index | – | – | – | – | 65.0%#4 | – | Epoch AI Benchmarking Hub |
Coding
| Benchmark | low | medium | high | xhigh | max | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| SciCode | 49.1%#106 | 52.9%#77 | 53.7%#65 | 57.3%#26 | 61.0%#5 | – | SciCode | |
| FrontierCode | – | – | – | 52.1%#5 | – | – | Cognition | |
| LMArena Coding | – | – | – | 1532#18 | – | – | LMArena | |
| LMArena WebDev | – | – | 1715#6 | 1774#3 | – | – | LMArena | |
| LiveBench Codingnot in index | – | – | – | 88.9%#4 | 91.4%#1 | – | LiveBench | |
| SciCode (AA)not in index | 49.1%#113 | 52.9%#77 | 53.7%#70 | 57.3%#27 | 61.0%#5 | – | Artificial Analysis | |
| IOIeffort not stated | – | – | – | – | – | 83.1%#9 | Vals AI | |
| Code Migrationnot in indexeffort not stated | – | – | – | – | – | 69.8%#1 | Vals AI | |
| Vibe Code Bench v1.1not in indexeffort not stated | – | – | – | – | – | 92.4%#1 | Vals AI | |
| CursorBenchnot in index | 35.8%#37 | 39.2%#29 | 47.8%#10 | 53.1%#5 | 55.5%#4 | – | Cursor | |
| ALE-Benchnot in index | – | – | 1819.1#10 | – | – | – | Epoch AI Benchmarking Hub | |
| FrontierSWEnot in index | – | – | – | – | 61.9%#3 | – | Epoch AI Benchmarking Hub | |
| Terminal-Bench 4.0 (AA)not in index | 20.7%#57 | 29.8%#43 | 43.9%#23 | 57.1%#6 | 63.6%#1 | – | Artificial Analysis |
Agents & tools
| Benchmark | low | medium | high | xhigh | max | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| Terminal-Bench | – | – | – | – | 61.8%#14 | – | Terminal-Bench | |
| APEX-Agents | – | 44.6%#35 | – | – | 75.5%#2 | – | Mercor | |
| LMArena Agent | – | – | – | – | 12#4 | – | LMArena | |
| LiveBench Agentic Codingnot in index | – | – | – | 39.3%#62 | 56.3%#27 | – | LiveBench | |
| Terminal-Bench 2.1 (Vals)effort not stated | – | – | – | – | – | 83.2%#6 | Vals AI | |
| GDPval-AA v2.1not in index | 33.9%#118 | 41.2%#92 | 52.5%#37 | 61.5%#6 | 67.0%#2 | – | Artificial Analysis | |
| Analyst Agent (AA)not in index | – | – | – | – | 57.5%#2 | – | Artificial Analysis | |
| CyberBenchnot in indexeffort not stated | – | – | – | – | – | 59.6%#30 | Vals AI | |
| Finance Agent v2not in indexeffort not stated | – | – | – | – | – | 58.1%#10 | Vals AI | |
| Harvey's Legal Agent Benchmarknot in indexeffort not stated | – | – | – | – | – | 2.9%#36 | Vals AI | |
| Legal Research Benchnot in indexeffort not stated | – | – | – | – | – | 48.1%#8 | Vals AI | |
| SREBenchnot in indexeffort not stated | – | – | – | – | – | 30.1%#6 | Vals AI | |
| Tax Agent Benchnot in indexeffort not stated | – | – | – | – | – | 73.4%#4 | Vals AI | |
| Terminal-Bench 4.0 (Vals)not in indexeffort not stated | – | – | – | – | – | 64.1%#2 | Vals AI | |
| Terminal-Bench Sciencenot in indexeffort not stated | – | – | – | – | – | 38.6%#5 | Vals AI | |
| AutomationBenchnot in index | – | – | – | 65.5%#12 | 71.8%#2 | – | Artificial Analysis | |
| GDP.pdfnot in index | – | – | – | 24.6%#16 | 25.8%#14 | – | Artificial Analysis | |
| Harvey LABnot in index | – | – | – | – | 2.8%#11 | – | Harvey | |
| AA-Briefcase v1.1not in index | – | – | – | 1752#4 | 1823#1 | – | Artificial Analysis |
Maths
| Benchmark | low | medium | high | xhigh | max | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| FrontierMath Tiers 1–3 | – | – | – | – | 88.8%#7 | – | Epoch AI Benchmarking Hub | |
| FrontierMath Tier 4 | – | – | – | – | 80.5%#13 | – | Epoch AI Benchmarking Hub | |
| OTIS Mock AIME | – | – | – | – | 100.0%#1 | – | Epoch AI Benchmarking Hub | |
| ProofBench | – | – | – | – | 100.0%#1 | – | Vals AI | |
| LiveBench Mathematicsnot in index | – | – | – | 96.7%#6 | 96.1%#10 | – | LiveBench | |
| LMArena Maths | – | – | – | 1512#8 | – | – | LMArena |
Knowledge
| Benchmark | low | medium | high | xhigh | max | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| SimpleQA Verified | – | – | – | – | 46.5%#34 | – | Epoch AI Benchmarking Hub | |
| LiveBench Data Analysisnot in index | – | – | – | 78.6%#21 | 59.5%#61 | – | LiveBench | |
| AA-Omniscience | 19.4#68 | 20.1#65 | 20.9#60 | 23.5#52 | 32.3#26 | – | Artificial Analysis | |
| BioMysteryBenchnot in indexeffort not stated | – | – | – | – | – | 81.1%#1 | Vals AI | |
| Excel Modeling Benchmarknot in indexeffort not stated | – | – | – | – | – | 75.7%#3 | Vals AI | |
| MedCodenot in indexeffort not stated | – | – | – | – | – | 52.9%#12 | Vals AI | |
| MedScribenot in indexeffort not stated | – | – | – | – | – | 91.1%#3 | Vals AI | |
| Public Benefits Benchnot in indexeffort not stated | – | – | – | – | – | 67.2%#12 | Vals AI | |
| AA-Omniscience: accuracynot in index | 46.3%#80 | 47.2%#75 | 52.0%#53 | 53.0%#48 | 54.0%#42 | – | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | 49.8%#110 | 48.8%#116 | 35.4%#181 | 37.1%#171 | 53.0%#100 | – | Artificial Analysis |
Instruction following
| Benchmark | low | medium | high | xhigh | max | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| LiveBench Languagenot in index | – | – | – | 83.4%#19 | 78.0%#38 | – | LiveBench | |
| LiveBench Instruction Followingnot in index | – | – | – | 70.5%#27 | 56.8%#61 | – | LiveBench | |
| LMArena Instruction Followingnot in index | – | – | – | 1489#13 | – | – | LMArena |
Human preference
| Benchmark | low | medium | high | xhigh | max | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| LMArena Text | – | – | – | 1476#31 | – | – | LMArena | |
| LMArena Creative Writingnot in index | – | – | – | 1463#24 | – | – | LMArena | |
| LMArena Multi-turnnot in index | – | – | – | 1484#31 | – | – | LMArena | |
| LMArena Longer Queriesnot in index | – | – | – | 1503#12 | – | – | LMArena |
Multimodal
| Benchmark | low | medium | high | xhigh | max | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| LMArena Vision | – | – | – | 1289#28 | – | – | LMArena | |
| SAGEnot in indexeffort not stated | – | – | – | – | – | 51.8%#11 | Vals AI |
Long context
| Benchmark | low | medium | high | xhigh | max | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| AA-LCR | 76.0%#140 | 76.3%#137 | 78.0%#114 | 79.7%#83 | 82.7%#33 | – | Artificial Analysis | |
| MLCRnot in index | – | – | – | – | 75.0%#1 | – | Artificial Analysis |
Composite
| Benchmark | low | medium | high | xhigh | max | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| Epoch Capabilities Indexnot in indexeffort not stated | – | – | – | – | – | 165.0#4 | Epoch AI Benchmarking Hub | |
| LiveBench | – | – | – | 77.8%#21 | 75.7%#35 | – | LiveBench | |
| AA Intelligence Index v4.3.2 | 35.9#81 | 40.8#56 | 46.8#28 | 51.9#10 | 56#2 | – | Artificial Analysis | |
| Vals Indexnot in indexeffort not stated | – | – | – | – | – | 67.0#2 | Vals AI |
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 | 121 tok/s | 2.26 s | $2.20 | $11.00 | 1M | – |
| Anthropic | 111 tok/s | 2.03 s | $2.00 | $10.00 | 1M | – |
| Azure | 110 tok/s | 1.19 s | $2.00 | $10.00 | 1M | – |
| Google Vertex | 100 tok/s | 0.57 s | $2.20 | $11.00 | 1M | – |
| Claude Platform on AWS | 98 tok/s | 2.64 s | $2.00 | $10.00 | 1M | – |
| Google Vertex | 95 tok/s | 1.80 s | $2.00 | $10.00 | 1M | – |
| Amazon Bedrock | 83 tok/s | 2.64 s | $2.00 | $10.00 | 1M | – |
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 $0.100 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0038 | $0.0032 | 8.2 min |
| Summarise a 30-page report | 12,000 / 600 | $0.030 | $0.013 | 8.2 min |
| Code edit | 6,000 / 1,500 | $0.027 | $0.018 | 8.3 min |
| Agentic coding session | 60,000 / 4,000 | $0.160 | $0.074 | 8.6 min |
| Structured extraction | 2,000 / 200 | $0.0060 | $0.0032 | 8.1 min |
See also
Data as of 11 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “Claude Sonnet 5.5: benchmarks, pricing, speed and rank”, https://www.benchleader.com/models/claude-sonnet-5-5, data as of 11 Oct 2026.