| Benchmark | Claude Opus 4.5 | o3 |
|---|---|---|
| AA Agentic Index | 59.6 | 36.1 |
| AA Intelligence | 41.9 | 31.1 |
| AA-LCR | 76 | 73.3 |
| AA-Omniscience | 14 | -15.3 |
| AIME 2025 | 92.8 | 89.2 |
| arc_agi_1 | 82 | 60.8 |
| ARC-AGI-2 | 37.6 | 6.5 |
| Artificial Analysis Coding Index | 47.8 | 38.4 |
| browsecomp | 37 | 49.7 |
| critpt | 4.6 | 1.1 |
| ERQA | 46.8 | 64 |
| Fortress | 13.6 | 16 |
| frontiermath_tier_4 | 4.2 | 2.1 |
| gdpval | 47.3 | 12.8 |
| GPQA Diamond | 87 | 83.3 |
| HLE | 30.8 | 20.3 |
| hmmt_feb_2025 | 92.9 | 77.5 |
| IFBench | 58 | 71.4 |
| LiveCodeBench | 87 | 84.7 |
| longbench_v2 | 64.4 | 58.8 |
| MathVista | 80.2 | 86.8 |
| metr_hcast | 73.4 | 64.7 |
| metr_re_bench | 40 | 15 |
| metr_swaa | 99.5 | 99.8 |
| MMLU-Pro | 90 | 85 |
| MMMU | 80.7 | 82.9 |
| MMMU-Pro | 74 | 76.4 |
| multichallenge | 59 | 60.4 |
| MultiNRC | 48.6 | 45.5 |
| OmniScience Accuracy | 46.6 | 38.6 |
| OmniScience Non-Hallucination | 39 | 12.9 |
| scicode | 50 | 41 |
| simplebench | 62 | 53.1 |
| simpleqa_verified | 41.8 | 53 |
| swe_bench_bash | 76.8 | 58.4 |
| SWE-bench Verified | 81.5 | 69.1 |
| Tau2 retail | 88.9 | 80.2 |
| Terminal-Bench Hard | 47 | 37.1 |
| VideoMMMU | 84.4 | 83.3 |
| τ²-Bench Telecom (AA run) | 89.5 | 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.