| Benchmark | Claude 4 Opus | Claude Opus 4.8 |
|---|---|---|
| AA Intelligence | 31.7 | 57.3 |
| AA-LCR | 40 | 73 |
| arc_agi_1 | 27 | 92.5 |
| ARC-AGI-2 | 8.6 | 72.1 |
| arena_vision | 1207 | 1280 |
| Artificial Analysis Coding Index | 34 | 74.3 |
| Fortress | 27.6 | 18.2 |
| frontiermath_tier_4 | 4.2 | 31.3 |
| GPQA Diamond | 79.6 | 93.6 |
| HLE | 12.3 | 57.9 |
| IFBench | 53.7 | 62.2 |
| livebench | 74.6 | 77.2 |
| longbench_v2 | 55.6 | 69.1 |
| scicode | 40.9 | 53.5 |
| simplebench | 58.8 | 64.8 |
| SWE-bench Verified | 79.4 | 88.6 |
| Terminal-Bench Hard | 31.1 | 58.3 |
| τ²-Bench Telecom (AA run) | 73.4 | 94.4 |
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.