| Benchmark | Claude Opus 4.8 | Kimi K2 (Reasoning) |
|---|---|---|
| AA Agentic Index | 49.4 | 47.9 |
| AA Intelligence | 57.3 | 33.5 |
| AA-LCR | 73 | 70.3 |
| AA-Omniscience | 28.8 | -21.4 |
| Artificial Analysis Coding Index | 74.3 | 34.8 |
| browsecomp | 88.5 | 60.2 |
| critpt | 20.9 | 2.6 |
| frontiermath_tier_4 | 31.3 | 0 |
| gdpval | 54.2 | 24.5 |
| GPQA Diamond | 93.6 | 84.5 |
| HealthBench | 59.3 | 58 |
| HLE | 57.9 | 23.9 |
| HLE (with tools) | 57.9 | 44.9 |
| hmmt_nov_2025 | 96.5 | 89.2 |
| IFBench | 62.2 | 68.1 |
| imo_answer_bench | 83.5 | 78.6 |
| longbench_v2 | 69.1 | 45.1 |
| OmniScience Accuracy | 48.8 | 30.9 |
| OmniScience Non-Hallucination | 60.7 | 25.8 |
| ResearchRubrics | 73.5 | 56.2 |
| scicode | 53.5 | 44.8 |
| simplebench | 64.8 | 39.6 |
| SWE-bench Multilingual | 84.4 | 61.1 |
| SWE-bench Verified | 88.6 | 71.3 |
| Terminal-Bench 2.0 | 74.6 | 35.7 |
| Terminal-Bench Hard | 58.3 | 31.1 |
| τ²-Bench Telecom (AA run) | 94.4 | 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.