| Benchmark | Gemini 2.5 Pro | GLM-5.2 Full Open Source |
|---|---|---|
| AA Agentic Index | 7.2 | 45.7 |
| AA Intelligence | 27 | 52.6 |
| AA-LCR | 66 | 76.7 |
| AA-Omniscience | -14.3 | 4.4 |
| arc_agi_1 | 41 | 77 |
| ARC-AGI-2 | 4.9 | 22.8 |
| arena_elo | 1446 | 1471 |
| arena_text_factuality | 1462 | 1460 |
| Artificial Analysis Coding Index | 46.7 | 68.8 |
| coding_arena_elo | 1224 | 1582 |
| critpt | 2.6 | 20.9 |
| gdpval | 8.5 | 50.3 |
| Global-MMLU-Lite | 88.6 | 89.2 |
| GPQA Diamond | 84.4 | 91.2 |
| HLE | 22.5 | 54.7 |
| HLE (with tools) | 28.4 | 54.7 |
| HMMT 2025 | 65.7 | 94.4 |
| hmmt_nov_2025 | 80 | 94.4 |
| IFBench | 49 | 73.3 |
| OmniScience Accuracy | 39 | 24.3 |
| OmniScience Non-Hallucination | 12.6 | 73.7 |
| scicode | 43 | 50.5 |
| simplebench | 62.4 | 58.8 |
| simpleqa_verified | 56 | 38.1 |
| TauBench V3 - Banking | 9.7 | 34.6 |
| Terminal-Bench 2.1 | 28.5 | 82.7 |
| Terminal-Bench Hard | 26.5 | 50.8 |
| τ²-Bench Telecom (AA run) | 54.1 | 99.1 |
| τ³-Bench | 9.3 | 26.8 |
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.