| Benchmark | Claude Opus 4.7 | DeepSeek-R1 |
|---|---|---|
| AA Agentic Index | 64.6 | 3.1 |
| AA Intelligence | 55 | 18.6 |
| AA-LCR | 75.3 | 56 |
| AA-Omniscience | 27.3 | -31.3 |
| arc_agi_1 | 92 | 21.2 |
| ARC-AGI-2 | 75.8 | 1.3 |
| Artificial Analysis Coding Index | 73.6 | 24.6 |
| browsecomp | 79.8 | 8.9 |
| critpt | 12 | 0.6 |
| gdpval | 58.6 | 1.5 |
| GPQA Diamond | 94.2 | 81 |
| harmbench | 17.5 | 54.6 |
| HLE | 54.7 | 17.7 |
| IFBench | 58.6 | 39 |
| OmniScience Accuracy | 48.9 | 30.7 |
| OmniScience Non-Hallucination | 57.7 | 10.5 |
| scicode | 54.5 | 35.7 |
| simplebench | 61.7 | 40.8 |
| simpleqa_verified | 50.6 | 27.4 |
| SWE-bench Multilingual | 80.5 | 30.5 |
| SWE-bench Verified | 87.6 | 57.6 |
| TauBench V3 - Banking | 34.6 | 6.4 |
| Terminal-Bench 2.1 | 83.1 | 19.1 |
| Terminal-Bench Hard | 54.5 | 6.1 |
| vectara_answer_rate | 98 | 97 |
| vectara_avg_summary_length | 149.1 | 93.5 |
| vectara_factual_consistency | 88 | 88.7 |
| vectara_hallucination_rate ↓ | 12 | 11.3 |
| τ²-Bench Telecom (AA run) | 88.6 | 11.4 |
| τ³-Bench | 28.9 | 6.4 |
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.