Anthropic
Claude Sonnet 4 benchmark results
17 results on 10 benchmarks from 5 sources.
Also listed as: claude-sonnet-4-20250514-thinking-32k · claude-sonnet-4-20250514 · claude-sonnet-4-20250514_59K · claude-sonnet-4-20250514_32K · claude-sonnet-4-20250514_16K · claude-sonnet-4-20250514_unknown · Claude Sonnet 4 (unknown thinking) · Claude Sonnet 4 (Thinking) · Claude 4 Sonnet · EPAM AI/Run Developer Agent v20250719 + Claude 4 Sonnet
All benchmark results
| Benchmarks | Score | Rank | Setting | Source type | Date |
|---|---|---|---|---|---|
| LMArena Text | 1402 | 158 | Overall, style control on; 34149 votes; 95% CI 1397–1406“claude-sonnet-4-20250514-thinking-32k” | Official leaderboard | Oct 8, 2026 |
| LMArena Text | 1391 | 173 | Overall, style control on; 39172 votes; 95% CI 1386–1395“claude-sonnet-4-20250514” | Official leaderboard | Oct 8, 2026 |
| LMArena Vision | 1206 | 82 | Overall, style control on; 1145 votes; 95% CI 1189–1223“claude-sonnet-4-20250514-thinking-32k” | Official leaderboard | Oct 2, 2026 |
| LMArena Vision | 1188 | 93 | Overall, style control on; 1807 votes; 95% CI 1174–1202“claude-sonnet-4-20250514” | Official leaderboard | Oct 2, 2026 |
| LiveCodeBench | 48.9% | 16 | Code generation, 182 problems dated 2025-01-01…2025-04-07“Claude-Sonnet-4 (Thinking)” | Official leaderboard | Apr 7, 2025 |
| GPQA Diamond | 79.2% | 119 | Epoch AI run, accuracy | Independent run | May 26, 2025 |
| SWE-bench Verified | 76.8% | 4 | agent: EPAM AI/Run Developer Agent“claude-sonnet-4-20250514” | Official leaderboard | Aug 4, 2025 |
| SWE-bench Verified | 74.8% | 9 | agent: Harness AI“claude-sonnet-4-20250514” | Official leaderboard | Jul 31, 2025 |
| SWE-bench Verified | 57.0% | 55 | agent: Artemis Agent v2“claude-sonnet-4-20250514” | Official leaderboard | Sep 24, 2025 |
| AIME (OTIS Mock AIME 2024–2025) | 68.9% | 131 | Epoch AI run, OTIS Mock AIME 2024–2025 | Independent run | May 23, 2025 |
| MATH Level 5 | 84.4% | 30 | Epoch AI run, accuracy | Independent run | May 22, 2025 |
| Humanity's Last Exam | 7.8% | 40 | Compiled by Epoch AI; effort: unknown thinking | Third-party compilation via Epoch AI | May 22, 2025 |
| OSWorld (Verified) | 43.9% | 5 | OSWorld; agent: claude-4-sonnet-20250514 (50 steps) | Official leaderboard via Epoch AI | Jul 27, 2025 |
| OSWorld (Verified) | 41.4% | 7 | OSWorld; agent: claude-4-sonnet-20250514 (100 steps) | Official leaderboard via Epoch AI | Jul 27, 2025 |
| OSWorld (Verified) | 31.2% | 11 | OSWorld; agent: claude-4-sonnet-20250514 (15 steps) | Official leaderboard via Epoch AI | Jul 27, 2025 |
| τ²-bench | 60.0% | 15 | airline domain, pass^1; submitted by Anthropic; voriginal tau-bench | Vendor-reported | May 22, 2025 |
| τ²-bench | 80.5% | 11 | retail domain, pass^1; submitted by Anthropic; voriginal tau-bench | Vendor-reported | May 22, 2025 |
FAQ
Where do these Claude Sonnet 4 scores come from?
From Epoch AI Benchmarking Hub, LiveCodeBench leaderboard, LMArena leaderboard dataset, SWE-bench leaderboards (swebench.com), and τ²-bench leaderboard submissions. Each row links to the original leaderboard, dataset or report, and vendor-reported numbers are labelled.