| Benchmark | GPT-4o | Kimi K2.5 |
|---|---|---|
| AA Agentic Index | 9.7 | 52.8 |
| AA Intelligence | 11.2 | 36 |
| AA-LCR | 55 | 73 |
| AA-Omniscience | -10.5 | -7.3 |
| AIME 2025 | 11.7 | 96.1 |
| arc_agi_1 | 4.5 | 65.3 |
| ARC-AGI-2 | 0 | 11.8 |
| arena_vision | 1162 | 1247 |
| Artificial Analysis Coding Index | 24.2 | 46.8 |
| critpt | 0 | 3.1 |
| Fortress | 47.2 | 41.1 |
| gdpval | 0 | 38.3 |
| GPQA Diamond | 70.1 | 87.9 |
| HLE | 5.3 | 50.2 |
| IFBench | 36 | 70.2 |
| LiveCodeBench v6 | 30.9 | 85 |
| longbench_v2 | 48.1 | 61 |
| mathvision | 30.4 | 84.2 |
| MathVista | 63.8 | 90.1 |
| MMLU-Pro | 74.7 | 87.1 |
| MMMU-Pro | 59.9 | 78.5 |
| multichallenge | 40.3 | 61.4 |
| MultiNRC | 12.4 | 35.2 |
| OCRBench | 806 | 92.3 |
| OmniScience Accuracy | 23.7 | 35.2 |
| OmniScience Non-Hallucination | 62.1 | 50 |
| OSWorld-Verified | 5 | 63.3 |
| scicode | 33.4 | 49 |
| SEAL VISTA | 34.9 | 41.9 |
| simplebench | 17.8 | 46.8 |
| swe_bench_bash | 21.6 | 70.8 |
| SWE-bench Verified | 38.8 | 76.8 |
| Terminal-Bench Hard | 8.3 | 34.8 |
| vectara_answer_rate | 93.8 | 92.2 |
| vectara_avg_summary_length | 86.6 | 112 |
| vectara_factual_consistency | 90.4 | 85.8 |
| vectara_hallucination_rate ↓ | 9.6 | 14.2 |
| Video-MME | 77.2 | 87.4 |
| VideoMMMU | 61.2 | 86.6 |
| τ²-Bench Telecom (AA run) | 28.9 | 95.9 |
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.