| Benchmark | DeepSeek-R1 | GLM-5.1 |
|---|---|---|
| AA Agentic Index | 3.1 | 66 |
| AA Intelligence | 18.6 | 41 |
| AA-LCR | 56 | 68 |
| AA-Omniscience | -31.3 | 0.8 |
| Artificial Analysis Coding Index | 24.6 | 55.8 |
| browsecomp | 8.9 | 79.3 |
| critpt | 0.6 | 4.6 |
| gdpval | 1.5 | 49.5 |
| GPQA Diamond | 81 | 86.8 |
| HLE | 17.7 | 52.3 |
| HMMT 2025 | 79.4 | 94 |
| IFBench | 39 | 76.3 |
| MMLU-Pro | 85 | 86 |
| OmniScience Accuracy | 30.7 | 25.2 |
| OmniScience Non-Hallucination | 10.5 | 70.1 |
| scicode | 35.7 | 43.8 |
| simplebench | 40.8 | 55.1 |
| simpleqa_verified | 27.4 | 38.1 |
| SWE-bench Multilingual | 30.5 | 73.3 |
| TauBench V3 - Banking | 6.4 | 13.6 |
| Terminal-Bench 2.1 | 19.1 | 63.5 |
| Terminal-Bench Hard | 6.1 | 43.2 |
| τ²-Bench Telecom (AA run) | 11.4 | 97.7 |
| τ³-Bench | 6.4 | 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.