| Benchmark | Gemini 2.5 Flash | GLM-5.2 Full Open Source |
|---|---|---|
| AA Agentic Index | 18.8 | 45.7 |
| AA Intelligence | 17.7 | 52.6 |
| AA-LCR | 65.7 | 76.7 |
| AA-Omniscience | -29.8 | 4.4 |
| arc_agi_1 | 33.3 | 77 |
| ARC-AGI-2 | 2.5 | 22.8 |
| arena_text_factuality | 1437 | 1460 |
| Artificial Analysis Coding Index | 22.2 | 68.8 |
| critpt | 1.4 | 20.9 |
| gdpval | 11.9 | 50.3 |
| Global-MMLU-Lite | 88.4 | 89.2 |
| GPQA Diamond | 82.8 | 91.2 |
| HLE | 12.1 | 54.7 |
| IFBench | 50.3 | 73.3 |
| OmniScience Accuracy | 26.1 | 24.3 |
| OmniScience Non-Hallucination | 24.7 | 73.7 |
| scicode | 35.9 | 50.5 |
| simplebench | 41.2 | 58.8 |
| Terminal-Bench Hard | 13.6 | 50.8 |
| τ²-Bench Telecom (AA run) | 31.6 | 99.1 |
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.