| Benchmark | Claude Opus 4.5 | o4-mini |
|---|---|---|
| AA Agentic Index | 59.6 | 36.1 |
| AA Intelligence | 41.9 | 26.1 |
| AA-LCR | 76 | 60 |
| AA-Omniscience | 14 | -35.7 |
| AIME 2025 | 92.8 | 92.7 |
| arc_agi_1 | 82 | 58.7 |
| ARC-AGI-2 | 37.6 | 6.1 |
| Artificial Analysis Coding Index | 47.8 | 25.6 |
| browsecomp | 37 | 51.5 |
| critpt | 4.6 | 0.6 |
| Fortress | 13.6 | 21.5 |
| frontiermath_tier_4 | 4.2 | 6.3 |
| gdpval | 47.3 | 25.4 |
| GPQA Diamond | 87 | 81.4 |
| HLE | 30.8 | 16.5 |
| IFBench | 58 | 68.7 |
| LiveCodeBench | 87 | 84.5 |
| MathVista | 80.2 | 84.3 |
| MMMU | 80.7 | 81.6 |
| MMMU-Pro | 74 | 69.2 |
| multichallenge | 59 | 44.9 |
| MultiNRC | 48.6 | 22.2 |
| OmniScience Accuracy | 46.6 | 24.8 |
| OmniScience Non-Hallucination | 39 | 19.5 |
| scicode | 50 | 46.5 |
| SEAL VISTA | 46.4 | 51.8 |
| simplebench | 62 | 38.7 |
| simpleqa_verified | 41.8 | 23.9 |
| swe_bench_bash | 76.8 | 45 |
| SWE-bench Verified | 81.5 | 68.1 |
| Terminal-Bench Hard | 47 | 15.2 |
| vectara_answer_rate | 98.7 | 99.2 |
| vectara_avg_summary_length | 114.5 | 130.9 |
| vectara_factual_consistency | 89.1 | 81.4 |
| vectara_hallucination_rate ↓ | 10.9 | 18.6 |
| τ²-Bench Telecom (AA run) | 89.5 | 55.6 |
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.