| Benchmark | Gemini 2.5 Pro | Kimi K2 (Reasoning) |
|---|---|---|
| AA Agentic Index | 7.2 | 47.9 |
| AA Intelligence | 27 | 33.5 |
| AA-LCR | 66 | 70.3 |
| AA-Omniscience | -14.3 | -21.4 |
| aider_polyglot | 82.2 | 59.1 |
| AIME 2025 | 88.3 | 94.5 |
| ArtifactsBench | 57.7 | 54.2 |
| Artificial Analysis Coding Index | 46.7 | 34.8 |
| browsecomp | 9.9 | 60.2 |
| browsecomp_zh | 32.2 | 62.3 |
| critpt | 2.6 | 2.6 |
| FinSearchComp-global | 42.6 | 29.5 |
| frontiermath_tier_4 | 4.2 | 0 |
| GAIA (text only) | 60.2 | 60.2 |
| gdpval | 8.5 | 24.5 |
| GPQA Diamond | 84.4 | 84.5 |
| HLE | 22.5 | 23.9 |
| HLE (with tools) | 28.4 | 44.9 |
| HMMT 2025 | 65.7 | 89.4 |
| hmmt_feb_2025 | 82.5 | 93.3 |
| hmmt_nov_2025 | 80 | 89.2 |
| IFBench | 49 | 68.1 |
| LiveCodeBench | 82.7 | 79.2 |
| MMLU-Pro | 86 | 84.6 |
| mrcr | 93 | 44.2 |
| multichallenge | 53.6 | 66.4 |
| OmniScience Accuracy | 39 | 30.9 |
| OmniScience Non-Hallucination | 12.6 | 25.8 |
| scicode | 43 | 44.8 |
| simplebench | 62.4 | 39.6 |
| swe_bench_bash | 53.6 | 63.4 |
| SWE-bench Verified | 67.2 | 71.3 |
| Terminal-Bench | 25.3 | 47.1 |
| Terminal-Bench Hard | 26.5 | 31.1 |
| theagentcompany | 39.3 | 30 |
| vectara_answer_rate | 99.1 | 98.6 |
| vectara_avg_summary_length | 106.4 | 59.2 |
| vectara_factual_consistency | 93 | 82.1 |
| vectara_hallucination_rate ↓ | 7 | 17.9 |
| xbench_deepsearch | 56 | 76 |
| τ²-Bench | 54 | 74.3 |
| τ²-Bench Telecom (AA run) | 54.1 | 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.