| Benchmark | GLM-5.1 | Kimi K3 |
|---|---|---|
| AA Agentic Index | 66 | 54.3 |
| AA Intelligence | 41 | 60 |
| AA-LCR | 68 | 82.7 |
| AA-Omniscience | 0.8 | 46 |
| aime_2026 | 95.8 | 96.7 |
| arena_elo | 1468 | 1489 |
| arena_text_factuality | 1458 | 1471 |
| Artificial Analysis Coding Index | 55.8 | 76.2 |
| browsecomp | 79.3 | 91.2 |
| coding_arena_elo | 1510 | 1674 |
| critpt | 4.6 | 23.4 |
| cybergym | 68.7 | 80 |
| DeepSWE | 18 | 67.5 |
| Finance Agent v2 | 44.8 | 54.4 |
| gdpval | 49.5 | 59.4 |
| GPQA Diamond | 86.8 | 93.5 |
| HLE | 52.3 | 56 |
| hmmt_feb_2026 | 89.4 | 97 |
| itbenchSre | 40.3 | 47.7 |
| MCP Atlas | 75.6 | 84.2 |
| OmniScience Accuracy | 25.2 | 47.6 |
| OmniScience Non-Hallucination | 70.1 | 49.1 |
| PostTrainBench | 20.1 | 36.6 |
| Program Bench | 50.9 | 77.8 |
| scicode | 43.8 | 58.7 |
| simplebench | 55.1 | 60.7 |
| simpleqa_verified | 38.1 | 42.7 |
| SWE-Marathon | 1 | 42 |
| TauBench V3 - Banking | 13.6 | 46 |
| Terminal-Bench 2.1 | 63.5 | 88.3 |
| toolathlon | 40.7 | 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.