| Benchmark | Gemini 3 Flash Preview | Kimi K3 |
|---|---|---|
| AA Agentic Index | 49.7 | 54.3 |
| AA Intelligence | 38.7 | 60 |
| AA-LCR | 73 | 82.7 |
| AA-Omniscience | 10.1 | 46 |
| aime_2026 | 96.7 | 96.7 |
| apexAgents | 27.7 | 41.3 |
| arc_agi_1 | 84.7 | 65.7 |
| ARC-AGI-2 | 33.6 | 12.4 |
| arena_elo | 1473 | 1489 |
| arena_text_factuality | 1452 | 1471 |
| Artificial Analysis Coding Index | 42.6 | 76.2 |
| charxiv_rq | 80.3 | 84.8 |
| coding_arena_elo | 1387 | 1674 |
| critpt | 8.6 | 23.4 |
| Finance Agent v2 | 42.6 | 54.4 |
| gdpval | 35.2 | 59.4 |
| GPQA Diamond | 90.4 | 93.5 |
| HLE | 43.5 | 56 |
| hmmt_feb_2026 | 89.4 | 97 |
| matharena_visual_math_overall | 85.8 | 91.3 |
| MCP Atlas | 62 | 84.2 |
| MMMU-Pro | 81.2 | 83.4 |
| OmniScience Accuracy | 53.4 | 47.6 |
| OmniScience Non-Hallucination | 7.6 | 49.1 |
| OSWorld-Verified | 65.1 | 84.8 |
| scicode | 50.6 | 58.7 |
| simplebench | 61.1 | 60.7 |
| simpleqa_verified | 67.4 | 42.7 |
| TauBench V3 - Banking | 20.8 | 46 |
| Terminal-Bench 2.1 | 58 | 88.3 |
| toolathlon | 49.4 | 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.