| Benchmark | GPT-5.2 | Kimi K3 |
|---|---|---|
| AA Agentic Index | 60.2 | 54.3 |
| AA Intelligence | 43.3 | 60 |
| AA-LCR | 79.3 | 82.7 |
| AA-Omniscience | 0.3 | 46 |
| aime_2026 | 98.3 | 96.7 |
| apexAgents | 23 | 41.3 |
| arc_agi_1 | 86.2 | 65.7 |
| ARC-AGI-2 | 52.9 | 12.4 |
| arena_elo | 1437 | 1489 |
| arena_text_factuality | 1442 | 1471 |
| Artificial Analysis Coding Index | 48.7 | 76.2 |
| browsecomp | 65.8 | 91.2 |
| charxiv_rq | 82.1 | 84.8 |
| coding_arena_elo | 1418 | 1674 |
| critpt | 11.6 | 23.4 |
| deepsearchqa_f1 | 71.3 | 95 |
| gdpval | 48.3 | 59.4 |
| GPQA Diamond | 92.4 | 93.5 |
| HLE | 45.5 | 56 |
| hmmt_feb_2026 | 97 | 97 |
| livebench_agentic_coding | 50.3 | 62.2 |
| livebench_coding | 76.1 | 81.5 |
| livebench_data_analysis | 78.2 | 78.7 |
| livebench_instruction_following | 61.8 | 71.4 |
| livebench_language | 79.8 | 85.5 |
| livebench_math | 93.2 | 84.4 |
| livebench_reasoning | 83.2 | 90.7 |
| matharena_visual_math_overall | 86.5 | 91.3 |
| mathvision | 83 | 97.8 |
| MCP Atlas | 68 | 84.2 |
| MMMU-Pro | 79.5 | 83.4 |
| MMVU | 80.8 | 82.1 |
| OmniScience Accuracy | 44.3 | 47.6 |
| OmniScience Non-Hallucination | 38.4 | 49.1 |
| scicode | 52.1 | 58.7 |
| simplebench | 45.8 | 60.7 |
| simpleqa_verified | 38.9 | 42.7 |
| TauBench V3 - Banking | 11.1 | 46 |
| toolathlon | 46.3 | 73.2 |
| Video-MME | 86 | 90 |
| WorldVQA | 28 | 51 |
| ZeroBench | 9 | 0.4 |
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.