| Benchmark | Claude Opus 4.7 | Kimi K2 (Reasoning) |
|---|---|---|
| AA Agentic Index | 64.6 | 47.9 |
| AA Intelligence | 55 | 33.5 |
| AA-LCR | 75.3 | 70.3 |
| AA-Omniscience | 27.3 | -21.4 |
| Artificial Analysis Coding Index | 73.6 | 34.8 |
| browsecomp | 79.8 | 60.2 |
| critpt | 12 | 2.6 |
| frontiermath_tier_4 | 22.9 | 0 |
| gdpval | 58.6 | 24.5 |
| GPQA Diamond | 94.2 | 84.5 |
| HLE | 54.7 | 23.9 |
| HLE (with tools) | 54.7 | 44.9 |
| IFBench | 58.6 | 68.1 |
| OmniScience Accuracy | 48.9 | 30.9 |
| OmniScience Non-Hallucination | 57.7 | 25.8 |
| scicode | 54.5 | 44.8 |
| simplebench | 61.7 | 39.6 |
| SWE-bench Multilingual | 80.5 | 61.1 |
| SWE-bench Verified | 87.6 | 71.3 |
| Terminal-Bench 2.0 | 69.4 | 35.7 |
| Terminal-Bench Hard | 54.5 | 31.1 |
| vectara_answer_rate | 98 | 98.6 |
| vectara_avg_summary_length | 149.1 | 59.2 |
| vectara_factual_consistency | 88 | 82.1 |
| vectara_hallucination_rate ↓ | 12 | 17.9 |
| τ²-Bench Telecom (AA run) | 88.6 | 93 |
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.