| Benchmark | Claude 4 Opus | Claude Opus 4.5 |
|---|---|---|
| AA Intelligence | 31.7 | 41.9 |
| AA-LCR | 40 | 76 |
| AIME 2025 | 75.5 | 92.8 |
| arc_agi_1 | 27 | 82 |
| ARC-AGI-2 | 8.6 | 37.6 |
| Artificial Analysis Coding Index | 34 | 47.8 |
| Fortress | 27.6 | 13.6 |
| frontiermath_tier_4 | 4.2 | 4.2 |
| GPQA Diamond | 79.6 | 87 |
| HLE | 12.3 | 30.8 |
| hmmt_feb_2025 | 60 | 92.9 |
| IFBench | 53.7 | 58 |
| livebench | 74.6 | 76 |
| LiveCodeBench | 70.4 | 87 |
| LiveCodeBench v6 | 47.4 | 84.8 |
| longbench_v2 | 55.6 | 64.4 |
| metr_hcast | 61.7 | 73.4 |
| metr_re_bench | 20 | 40 |
| metr_swaa | 99.8 | 99.5 |
| mmlu_redux | 94.2 | 95.6 |
| MMLU-Pro | 86.6 | 90 |
| mmmlu | 88.8 | 90.8 |
| MMMU | 73.7 | 80.7 |
| multichallenge | 58.6 | 59 |
| MultiNRC | 33.9 | 48.6 |
| scicode | 40.9 | 50 |
| SEAL VISTA | 47 | 46.4 |
| simplebench | 58.8 | 62 |
| supergpqa | 56.5 | 70.6 |
| swe_bench_bash | 67.6 | 76.8 |
| SWE-bench Verified | 79.4 | 81.5 |
| Tau2 retail | 81.8 | 88.9 |
| Terminal-Bench Hard | 31.1 | 47 |
| vectara_answer_rate | 91 | 98.7 |
| vectara_avg_summary_length | 123.2 | 114.5 |
| vectara_factual_consistency | 88 | 89.1 |
| vectara_hallucination_rate ↓ | 12 | 10.9 |
| τ²-Bench | 57 | 98.2 |
| τ²-Bench Telecom (AA run) | 73.4 | 89.5 |
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.