| Benchmark | Claude Opus 4.5 | DeepSeek-R1 |
|---|---|---|
| AA Agentic Index | 59.6 | 3.1 |
| AA Intelligence | 41.9 | 18.6 |
| AA-LCR | 76 | 56 |
| AA-Omniscience | 14 | -31.3 |
| AIME 2025 | 92.8 | 87.5 |
| arc_agi_1 | 82 | 21.2 |
| ARC-AGI-2 | 37.6 | 1.3 |
| Artificial Analysis Coding Index | 47.8 | 24.6 |
| browsecomp | 37 | 8.9 |
| browsecomp_zh | 62.4 | 35.7 |
| C-Eval | 92.2 | 91.8 |
| critpt | 4.6 | 0.6 |
| Fortress | 13.6 | 74.4 |
| gdpval | 47.3 | 1.5 |
| GPQA Diamond | 87 | 81 |
| HLE | 30.8 | 17.7 |
| hmmt_feb_2025 | 92.9 | 76.7 |
| IFBench | 58 | 39 |
| LiveCodeBench | 87 | 84.4 |
| longbench_v2 | 64.4 | 58.3 |
| mmlu_redux | 95.6 | 93.4 |
| MMLU-Pro | 90 | 85 |
| multichallenge | 59 | 45 |
| MultiNRC | 48.6 | 27.6 |
| OmniScience Accuracy | 46.6 | 30.7 |
| OmniScience Non-Hallucination | 39 | 10.5 |
| scicode | 50 | 35.7 |
| simplebench | 62 | 40.8 |
| simpleqa_verified | 41.8 | 27.4 |
| SWE-bench Multilingual | 77.5 | 30.5 |
| SWE-bench Verified | 81.5 | 57.6 |
| Terminal-Bench Hard | 47 | 6.1 |
| vectara_answer_rate | 98.7 | 97 |
| vectara_avg_summary_length | 114.5 | 93.5 |
| vectara_factual_consistency | 89.1 | 88.7 |
| vectara_hallucination_rate ↓ | 10.9 | 11.3 |
| τ²-Bench Telecom (AA run) | 89.5 | 11.4 |
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.