| Benchmark | Kimi K3 | Qwen3.7 Plus Preview |
|---|---|---|
| AA Agentic Index | 54.3 | 20.8 |
| AA Intelligence | 60 | 39.4 |
| AA-LCR | 82.7 | 69 |
| AA-Omniscience | 46 | 2.4 |
| Agents' Last Exam | 28.3 | 33.6 |
| apexAgents | 41.3 | 22.4 |
| arena_elo | 1489 | 1458 |
| arena_text_factuality | 1471 | 1454 |
| Artificial Analysis Coding Index | 76.2 | 55.9 |
| baby_vision | 85.7 | 70.4 |
| charxiv_rq | 84.8 | 85.9 |
| critpt | 23.4 | 9.1 |
| DeepSWE 1.1 | 69 | 14.2 |
| Finance Agent v2 | 54.4 | 38.2 |
| gdpval | 59.4 | 22.2 |
| GPQA Diamond | 93.5 | 90.3 |
| HLE | 56 | 35.6 |
| hmmt_feb_2026 | 97 | 92.9 |
| JobBench | 54.3 | 27.6 |
| mathvision | 97.8 | 90.3 |
| MCP Atlas | 84.2 | 73.2 |
| MMMU-Pro | 83.4 | 80.5 |
| OmniScience Accuracy | 47.6 | 22.5 |
| OmniScience Non-Hallucination | 49.1 | 74.5 |
| OSWorld-Verified | 84.8 | 73.3 |
| scicode | 58.7 | 51.3 |
| TauBench V3 - Banking | 46 | 17.9 |
| Terminal-Bench 2.1 | 88.3 | 61 |
| Video-MME | 90 | 88 |
| WorldVQA | 51 | 61.1 |
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.