Claude Sonnet 5
Category scores
Scores per capability area with the model's rank among ranked models in that category.
- 1Multimodal#1582.0
- 2Reasoning81.3
- 3Knowledge#3481.2
- 4Agentic & tool use#1577.5
- 5Math69.4
- 6Coding#2662.8
Benchmark results (25)
As reported in system cards, papers and public evaluations. Position compares this model with every other model reporting the same benchmark.
Agentic & tool use
| Terminal-Bench 2.1 | 80.4 | #29 of 68 |
| BrowseComp | 84.7 | #13 of 50 |
| MazeBench | 0 | #20 of 42 |
| OSWorld-Verified | 81.2 | #8 of 35 |
| HLE w/ tools | 57.4 | #8 of 27 |
| ApprenticeBench | 16 | #10 of 18 |
| Terminal-Bench 3.0 | 14.6 | #9 of 13 |
Multimodal
| Design Arena Website | 1274 | #24 of 91 |
| CharXiv | 88.3 | #12 of 42 |
| ParseBench | 62.13 | #26 of 37 |
| CharXiv w/o tools | 77.0 | #14 of 14 |
| SWE-bench Multimodal | 28.1 | #7 of 7 |
Coding
| SWE-bench Verified | 85.2 | #8 of 98 |
| SWE-bench Pro | 63.2 | #18 of 69 |
| FrontierCode 1.1 Main | 42.7 | #11 of 16 |
| VulcanBench CII v1 | 89.2 | #2 of 3 |
Knowledge
| HLE | 57.4 | #7 of 69 |
| HLE w/o tools | 43.2 | #17 of 43 |
| LABBench2 | 80.1 | #7 of 7 |
| HLE-Verified | 31.0 | #6 of 6 |
Math
| FrontierMath v2 (Tiers 1-3) | 65.614 | #30 of 64 |
| FrontierMath v2 (Tier 4) | 29.268 | #27 of 51 |
Multilingual
| SWE Multilingual | 78.3 | #15 of 46 |
Pricing by source
USD per 1M tokens.
| Source | Input | Output | Cached in |
|---|---|---|---|
| ✓ Official (Anthropic) checked Oct 17, 2026 · Claude Sonnet 5 | $2.00 | $10.00 | $0.20 |
| BenchLM list price | $2.00 | $10.00 | $0.20 |
| OpenRouter anthropic/claude-sonnet-5 | $2.00 | $10.00 | $0.20 |
Anthropic's Claude API pricing page, read September 25, 2026, lists Claude Sonnet 5 at $2 input / $0.20 cache hit / $10 output per million tokens; five-minute and one-hour cache writes are $2.50 and $4, and Batch API input and output are $1 and $5. A footnote on the $2/$10 row states that this pricing, announced at launch as introductory pricing through August 31, 2026, is now the standard price, and that the previously scheduled increase to $3/$15 on September 1, 2026 will not occur. Anthropic's launch post carries an August 10, 2026 edit saying the introductory pricing is now permanent (https://www.anthropic.com/news/claude-sonnet-5). The rate has applied since the June 30, 2026 launch. 1M-token context window.
Arena ratings
Human-preference ratings from LMArena. Best-rated variant shown.
| Text(high) | 1442 | #72 · 49.6K votes |
| WebDev (coding)(High) | 1540 | #47 · 15.8K votes |
| Vision(High) | 1273 | #41 · 14K votes |
| Documents(high) | 1466 | #17 · 7.4K votes |
Speed
Measured by us through OpenRouter (median of trials).
Not measured yet.
Related models
Same family
- Claude Sonnet 4.655.9
- Claude Opus 4.870.0
Similar score
- Gemini 3.1 Pro64.7 · $4.50
- MiMo-V2.6-Flash64.6 · $0.17
- GPT-5.6 Luna64.4 · $0.45
- Claude Opus 4.765.5 · $10.00
- DeepSeek V4.1 Flash 101765.7 · $0.26