| Benchmark | Claude 4 Opus | Gemini 1.5 Pro |
|---|---|---|
| AA Intelligence | 31.7 | 10 |
| aider_polyglot | 70.7 | 16.9 |
| AIR-Bench 2024 | 85.7 | 82.8 |
| anthropic_red_team | 98.9 | 99.9 |
| ARC-AGI-2 | 8.6 | 0.8 |
| arena_vision | 1207 | 1179 |
| Artificial Analysis Coding Index | 34 | 23.6 |
| bbq | 99.3 | 94.5 |
| Fortress | 27.6 | 53.9 |
| GPQA Diamond | 79.6 | 59.1 |
| harmbench | 91.7 | 79.9 |
| HLE | 12.3 | 4.9 |
| ifeval | 87.4 | 89.4 |
| LiveCodeBench | 70.4 | 30.5 |
| mmlu | 92.9 | 86.8 |
| MMLU-Pro | 86.6 | 75.8 |
| MMMU | 73.7 | 68.4 |
| scicode | 40.9 | 29.5 |
| SEAL VISTA | 47 | 37.1 |
| simple_safety_tests | 100 | 97.5 |
| simplebench | 58.8 | 27.1 |
| simpleqa | 22.8 | 23.4 |
| SWE-bench Verified | 79.4 | 34.2 |
| xstest | 97 | 98.8 |
Best tracked score per model per benchmark (default configuration; source-attributed). ↓ marks lower-is-better metrics. Open either model for its full surface, provenance and pricing.