| Benchmark | Gemini 2.5 Pro | GLM-5.1 |
|---|---|---|
| AA Agentic Index | 7.2 | 66 |
| AA Intelligence | 27 | 41 |
| AA-LCR | 66 | 68 |
| AA-Omniscience | -14.3 | 0.8 |
| arena_elo | 1446 | 1468 |
| arena_text_factuality | 1462 | 1458 |
| Artificial Analysis Coding Index | 46.7 | 55.8 |
| browsecomp | 9.9 | 79.3 |
| coding_arena_elo | 1224 | 1510 |
| critpt | 2.6 | 4.6 |
| frontiermath_tier_4 | 4.2 | 12.5 |
| gdpval | 8.5 | 49.5 |
| GPQA Diamond | 84.4 | 86.8 |
| HLE | 22.5 | 52.3 |
| HLE (with tools) | 28.4 | 52.3 |
| HMMT 2025 | 65.7 | 94 |
| hmmt_nov_2025 | 80 | 94 |
| IFBench | 49 | 76.3 |
| MMLU-Pro | 86 | 86 |
| OmniScience Accuracy | 39 | 25.2 |
| OmniScience Non-Hallucination | 12.6 | 70.1 |
| scicode | 43 | 43.8 |
| simplebench | 62.4 | 55.1 |
| simpleqa_verified | 56 | 38.1 |
| TauBench V3 - Banking | 9.7 | 13.6 |
| Terminal-Bench 2.1 | 28.5 | 63.5 |
| Terminal-Bench Hard | 26.5 | 43.2 |
| τ²-Bench Telecom (AA run) | 54.1 | 97.7 |
| τ³-Bench | 9.3 | 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.