| Benchmark | Gemini 3.5 Flash | Kimi K3 |
|---|---|---|
| AA Agentic Index | 70.4 | 54.3 |
| AA Intelligence | 52 | 60 |
| AA-LCR | 81 | 82.7 |
| AA-Omniscience | 21.2 | 46 |
| aime_2026 | 95 | 96.7 |
| apexAgents | 47.1 | 41.3 |
| arc_agi_1 | 92.5 | 65.7 |
| ARC-AGI-2 | 72.1 | 12.4 |
| arena_elo | 1477 | 1489 |
| arena_text_factuality | 1466 | 1471 |
| Artificial Analysis Coding Index | 70.1 | 76.2 |
| charxiv_rq | 84.2 | 84.8 |
| coding_arena_elo | 1490 | 1674 |
| critpt | 13.1 | 23.4 |
| DeepSWE | 37 | 67.5 |
| DeepSWE 1.1 | 37 | 69 |
| Finance Agent v2 | 57.9 | 54.4 |
| gdpval | 57.8 | 59.4 |
| GDPval-AA v2 | 1357 | 1687 |
| GPQA Diamond | 92.2 | 93.5 |
| HLE | 42.7 | 56 |
| hmmt_feb_2026 | 95.5 | 97 |
| itbenchSre | 40.3 | 47.7 |
| livebench_agentic_coding | 49 | 62.2 |
| livebench_coding | 78.2 | 81.5 |
| livebench_data_analysis | 64.9 | 78.7 |
| livebench_instruction_following | 75.6 | 71.4 |
| livebench_language | 84.6 | 85.5 |
| livebench_math | 88.2 | 84.4 |
| livebench_reasoning | 82 | 90.7 |
| matharena_visual_math_overall | 89.9 | 91.3 |
| MCP Atlas | 83.6 | 84.2 |
| MMMU-Pro | 84.3 | 83.4 |
| OmniScience Accuracy | 51.4 | 47.6 |
| OmniScience Non-Hallucination | 38.2 | 49.1 |
| OSWorld-Verified | 78.4 | 84.8 |
| scicode | 53.1 | 58.7 |
| simplebench | 76.7 | 60.7 |
| simpleqa_verified | 68.4 | 42.7 |
| TauBench V3 - Banking | 32.2 | 46 |
| Terminal-Bench 2.1 | 78.7 | 88.3 |
| toolathlon | 56.5 | 73.2 |
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.