| Benchmark | DeepSeek-V3 | GLM-5.1 |
|---|---|---|
| AA Agentic Index | 1.6 | 66 |
| AA Intelligence | 15.4 | 41 |
| AA-LCR | 41.3 | 68 |
| AA-Omniscience | -37.6 | 0.8 |
| Artificial Analysis Coding Index | 23 | 55.8 |
| critpt | 0 | 4.6 |
| gdpval | 0 | 49.5 |
| GPQA Diamond | 68.4 | 86.8 |
| HLE | 5.2 | 52.3 |
| HMMT 2025 | 27.5 | 94 |
| IFBench | 41 | 76.3 |
| livebench | 72.4 | 70.2 |
| MMLU-Pro | 81.2 | 86 |
| OmniScience Accuracy | 25.4 | 25.2 |
| OmniScience Non-Hallucination | 23.3 | 70.1 |
| scicode | 35.8 | 43.8 |
| simplebench | 27.2 | 55.1 |
| TauBench V3 - Banking | 4.7 | 13.6 |
| Terminal-Bench 2.1 | 16.9 | 63.5 |
| Terminal-Bench Hard | 15.2 | 43.2 |
| τ²-Bench Telecom (AA run) | 47.1 | 97.7 |
| τ³-Bench | 4.7 | 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.