AnthropicClaude Sonnet
Claude Sonnet 4
Anthropic · Claude Sonnet. Released May 22, 2025.
Claude Sonnet 4 is a model from Anthropic in the Claude Sonnet family, released May 22, 2025. evals.report tracks 34 reported Claude Sonnet 4 benchmark scores across ARC-AGI-1, ARC-AGI-2, SWE-bench Pro, GPQA Diamond, FrontierMath, AIME (OTIS Mock), LMArena, WeirdML, and 26 more — each shown with its benchmark, metric, source status, and date, and never combined into a single ranking.
34 results
Benchmark results 34
Compare this model| Benchmark | Category | Score | Metric | Status | Date | |
|---|---|---|---|---|---|---|
| ARC-AGI-1 | Reasoning | 40% | accuracy | Official | May 22, 2025 | Details |
| ARC-AGI-2 | Reasoning | 5.93% | accuracy | Official | May 22, 2025 | Details |
| SWE-bench Pro | Coding | 42.70% | % resolved | Official | May 22, 2025 | Details |
| GPQA Diamond | Reasoning | 78.3% | accuracy | Official | May 22, 2025 | Details |
| FrontierMath | Reasoning | 4.14% | accuracy | Official | May 22, 2025 | Details |
| AIME (OTIS Mock) | Reasoning | 71.1% | accuracy | Official | May 22, 2025 | Details |
| LMArena | Chat preference | 1337 | source-defined rating | Official | May 22, 2025 | Details |
| WeirdML | Coding | 46.1% | average accuracy | Official | May 22, 2025 | Details |
| MCP Atlas | Tool use | 35.6% | pass rate | Official | May 22, 2025 | Details |
| Artificial Analysis Intelligence Index | Reasoning | 33 | Index | Unverified | May 22, 2025 | Details |
| Epoch Capabilities Index | Reasoning | 142.6 | Index | Official | May 22, 2025 | Details |
| Aider Polyglot | Coding | 61.3% | % correct | Official | May 22, 2025 | Details |
| MMLU-Pro | Reasoning | 84.2% | accuracy | Verified | May 22, 2025 | Details |
| GDPval | Agents | 1133 | Elo | Official | May 22, 2025 | Details |
| LiveCodeBench | Coding | 44.9% | Pass@1 | Unverified | May 22, 2025 | Details |
| METR Task-Completion Time Horizons | Agents | 75 min | 50% time horizon | Official | May 22, 2025 | Details |
| SciCode | Coding | 40.0% | accuracy | Unverified | May 22, 2025 | Details |
| MMMU (Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark) | Multimodal | 74.4% | accuracy | Unverified | May 22, 2025 | Details |
| IFBench | Reasoning | 42.3% | accuracy | Official | May 22, 2025 | Details |
| MultiChallenge | Reasoning | 57.11% | accuracy | Verified | May 22, 2025 | Details |
| EQ-Bench Creative Writing v3 | Chat preference | 1514 | Elo | Verified | May 22, 2025 | Details |
| Design Arena | Chat preference | 1196 | Elo | Verified | May 22, 2025 | Details |
| MASK (Model Alignment between Statements and Knowledge) | Other | 95.33 | Honesty score | Verified | May 22, 2025 | Details |
| MCP-Universe | Tool use | 32.90% | Overall Success Rate | Verified | May 22, 2025 | Details |
| OCRBench v2 | Multimodal | 42.4 | accuracy | Verified | May 22, 2025 | Details |
| FACTS Grounding | Reasoning | 56.1% | Grounding accuracy | Verified | May 22, 2025 | Details |
| SWE-bench Multimodal | Coding | 35.59% | % resolved | Verified | May 22, 2025 | Details |
| EnigmaEval | Reasoning | 3.12% | accuracy | Verified | May 22, 2025 | Details |
| ZeroBench | Multimodal | 2.0% (pass@1) | accuracy | Verified | May 22, 2025 | Details |
| IMO-Bench | Reasoning | 23.0% | accuracy | Verified | May 22, 2025 | Details |
| Vectara Hallucination Leaderboard | Other | 10.3% | Hallucination Rate | Official | May 22, 2025 | Details |
| Online-Mind2Web | Agents | 40.00% | Task success rate | Verified | May 22, 2025 | Details |
| GSO: Software Optimization Benchmark for SWE-Agents | Coding | 4.9% | Opt@1 | Official | May 22, 2025 | Details |
| MultiNRC | Reasoning | 18.39% | accuracy | Official | May 22, 2025 | Details |