| Benchmark | Gemini 3.6 Flash | GLM-5.1 |
|---|---|---|
| AA Agentic Index | 40.5 | 66 |
| AA Intelligence | 52 | 41 |
| AA-LCR | 79 | 68 |
| AA-Omniscience | 22.1 | 0.8 |
| aime_2026 | 96.7 | 95.8 |
| arena_elo | 1481 | 1468 |
| arena_text_factuality | 1470 | 1458 |
| Artificial Analysis Coding Index | 69.2 | 55.8 |
| coding_arena_elo | 1538 | 1510 |
| critpt | 10.6 | 4.6 |
| DeepSWE | 49 | 18 |
| gdpval | 46.1 | 49.5 |
| GPQA Diamond | 92.8 | 86.8 |
| HLE | 40.8 | 52.3 |
| hmmt_feb_2026 | 89.4 | 89.4 |
| OmniScience Accuracy | 50 | 25.2 |
| OmniScience Non-Hallucination | 44.4 | 70.1 |
| scicode | 52.7 | 43.8 |
| simpleqa_verified | 68.7 | 38.1 |
| SWE-bench Pro | 58.7 | 58.4 |
| TauBench V3 - Banking | 29.9 | 13.6 |
| Terminal-Bench 2.1 | 78 | 63.5 |
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.