| Benchmark | Gemini 3.5 Flash | GLM-5.1 |
|---|---|---|
| AA Agentic Index | 70.4 | 66 |
| AA Intelligence | 52 | 41 |
| AA-LCR | 81 | 68 |
| AA-Omniscience | 21.2 | 0.8 |
| aime_2026 | 95 | 95.8 |
| arena_elo | 1477 | 1468 |
| arena_text_factuality | 1466 | 1458 |
| Artificial Analysis Coding Index | 70.1 | 55.8 |
| coding_arena_elo | 1490 | 1510 |
| critpt | 13.1 | 4.6 |
| DeepSWE | 37 | 18 |
| Finance Agent v2 | 57.9 | 44.8 |
| frontiermath_tier_4 | 14.6 | 12.5 |
| gdpval | 57.8 | 49.5 |
| GDPval-AA (Elo) | 1656 | 1535 |
| GPQA Diamond | 92.2 | 86.8 |
| HLE | 42.7 | 52.3 |
| hmmt_feb_2026 | 95.5 | 89.4 |
| IFBench | 76.3 | 76.3 |
| itbenchSre | 40.3 | 40.3 |
| livebench | 75 | 70.2 |
| MCP Atlas | 83.6 | 75.6 |
| OmniScience Accuracy | 51.4 | 25.2 |
| OmniScience Non-Hallucination | 38.2 | 70.1 |
| scicode | 53.1 | 43.8 |
| simplebench | 76.7 | 55.1 |
| simpleqa_verified | 68.4 | 38.1 |
| SWE-bench Pro | 55.1 | 58.4 |
| TauBench V3 - Banking | 32.2 | 13.6 |
| Terminal-Bench 2.0 | 76.2 | 69 |
| Terminal-Bench 2.1 | 78.7 | 63.5 |
| Terminal-Bench Hard | 46.2 | 43.2 |
| toolathlon | 56.5 | 40.7 |
| τ²-Bench Telecom (AA run) | 95.6 | 97.7 |
| τ³-Bench | 25.4 | 70.6 |
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.