| Benchmark | GPT-4o | Kimi K2 (Reasoning) |
|---|---|---|
| AA Agentic Index | 9.7 | 47.9 |
| AA Intelligence | 11.2 | 33.5 |
| AA-LCR | 55 | 70.3 |
| AA-Omniscience | -10.5 | -21.4 |
| aider_polyglot | 30.7 | 59.1 |
| AIME 2025 | 11.7 | 94.5 |
| Artificial Analysis Coding Index | 24.2 | 34.8 |
| critpt | 0 | 2.6 |
| gdpval | 0 | 24.5 |
| GPQA Diamond | 70.1 | 84.5 |
| HLE | 5.3 | 23.9 |
| IFBench | 36 | 68.1 |
| LiveCodeBench | 38.3 | 79.2 |
| LiveCodeBench v6 | 30.9 | 83.1 |
| longbench_v2 | 48.1 | 45.1 |
| mmlu_redux | 88 | 94.4 |
| MMLU-Pro | 74.7 | 84.6 |
| multichallenge | 40.3 | 66.4 |
| OmniScience Accuracy | 23.7 | 30.9 |
| OmniScience Non-Hallucination | 62.1 | 25.8 |
| scicode | 33.4 | 44.8 |
| simplebench | 17.8 | 39.6 |
| swe_bench_bash | 21.6 | 63.4 |
| SWE-bench Verified | 38.8 | 71.3 |
| Terminal-Bench Hard | 8.3 | 31.1 |
| vectara_answer_rate | 93.8 | 98.6 |
| vectara_avg_summary_length | 86.6 | 59.2 |
| vectara_factual_consistency | 90.4 | 82.1 |
| vectara_hallucination_rate ↓ | 9.6 | 17.9 |
| τ²-Bench Telecom (AA run) | 28.9 | 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.