| Benchmark | Kimi K2.5 | o4-mini |
|---|---|---|
| AA Agentic Index | 52.8 | 36.1 |
| AA Intelligence | 36 | 26.1 |
| AA-LCR | 73 | 60 |
| AA-Omniscience | -7.3 | -35.7 |
| AIME 2025 | 96.1 | 92.7 |
| arc_agi_1 | 65.3 | 58.7 |
| ARC-AGI-2 | 11.8 | 6.1 |
| arena_vision | 1247 | 1201 |
| Artificial Analysis Coding Index | 46.8 | 25.6 |
| browsecomp | 74.9 | 51.5 |
| critpt | 3.1 | 0.6 |
| Fortress | 41.1 | 21.5 |
| frontiermath_tier_4 | 4.2 | 6.3 |
| gdpval | 38.3 | 25.4 |
| GPQA Diamond | 87.9 | 81.4 |
| HLE | 50.2 | 16.5 |
| IFBench | 70.2 | 68.7 |
| MathVista | 90.1 | 84.3 |
| MMMU-Pro | 78.5 | 69.2 |
| multichallenge | 61.4 | 44.9 |
| MultiNRC | 35.2 | 22.2 |
| OmniScience Accuracy | 35.2 | 24.8 |
| OmniScience Non-Hallucination | 50 | 19.5 |
| scicode | 49 | 46.5 |
| SEAL VISTA | 41.9 | 51.8 |
| simplebench | 46.8 | 38.7 |
| simpleqa_verified | 36.9 | 23.9 |
| strongreject | 99.5 | 96 |
| swe_bench_bash | 70.8 | 45 |
| SWE-bench Verified | 76.8 | 68.1 |
| Terminal-Bench Hard | 34.8 | 15.2 |
| vectara_answer_rate | 92.2 | 99.2 |
| vectara_avg_summary_length | 112 | 130.9 |
| vectara_factual_consistency | 85.8 | 81.4 |
| vectara_hallucination_rate ↓ | 14.2 | 18.6 |
| τ²-Bench Telecom (AA run) | 95.9 | 55.6 |
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.