| Benchmark | Claude Opus 4.6 | o3 |
|---|---|---|
| AA Agentic Index | 67.6 | 36.1 |
| AA Intelligence | 44.9 | 31.1 |
| AA-LCR | 74.3 | 73.3 |
| AA-Omniscience | 13.7 | -15.3 |
| AIME 2025 | 99.8 | 89.2 |
| arc_agi_1 | 93 | 60.8 |
| ARC-AGI-2 | 68.8 | 6.5 |
| arena_vision | 1300 | 1217 |
| Artificial Analysis Coding Index | 48.1 | 38.4 |
| browsecomp | 86.8 | 49.7 |
| critpt | 12.6 | 1.1 |
| ERQA | 40.8 | 64 |
| Fortress | 20.5 | 16 |
| frontiermath_tier_4 | 22.9 | 2.1 |
| gdpval | 55.9 | 12.8 |
| GPQA Diamond | 91.3 | 83.3 |
| HLE | 53.1 | 20.3 |
| IFBench | 62.5 | 71.4 |
| LiveCodeBench | 88.8 | 84.7 |
| metr_hcast | 77 | 64.7 |
| metr_re_bench | 97.4 | 15 |
| metr_swaa | 98.5 | 99.8 |
| MMLU-Pro | 89.1 | 85 |
| MMMU-Pro | 77.3 | 76.4 |
| multichallenge | 56 | 60.4 |
| MultiNRC | 57.1 | 45.5 |
| OmniScience Accuracy | 47 | 38.6 |
| OmniScience Non-Hallucination | 37.2 | 12.9 |
| scicode | 52 | 41 |
| simplebench | 67.6 | 53.1 |
| simpleqa_verified | 46.5 | 53 |
| swe_bench_bash | 75.6 | 58.4 |
| SWE-bench Verified | 80.8 | 69.1 |
| Tau2 retail | 91.9 | 80.2 |
| Terminal-Bench Hard | 65.4 | 37.1 |
| VisualToolBench | 27.5 | 13.7 |
| τ²-Bench Telecom (AA run) | 99.3 | 80.7 |
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.