| Benchmark | DeepSeek-V4-Flash | GLM-5.1 |
|---|---|---|
| AA Agentic Index | 61.3 | 66 |
| AA Intelligence | 51.8 | 41 |
| AA-LCR | 74.3 | 68 |
| AA-Omniscience | -14.3 | 0.8 |
| aime_2026 | 95.8 | 95.8 |
| Apex (Pass@1) | 33 | 11.5 |
| Apex Shortlist (Pass@1) | 85.7 | 72.4 |
| arena_elo | 1439 | 1468 |
| arena_text_factuality | 1434 | 1458 |
| Artificial Analysis Coding Index | 69.1 | 55.8 |
| browsecomp | 73.2 | 79.3 |
| coding_arena_elo | 1431 | 1510 |
| critpt | 16.6 | 4.6 |
| cybergym | 76.7 | 68.7 |
| DeepSWE | 54.4 | 18 |
| gdpval | 52.9 | 49.5 |
| GDPval-AA (Elo) | 1395 | 1535 |
| GPQA Diamond | 90.8 | 86.8 |
| HLE | 38.6 | 52.3 |
| HLE (with tools) | 45.1 | 52.3 |
| hmmt_feb_2026 | 94.8 | 89.4 |
| IFBench | 79.2 | 76.3 |
| imo_answer_bench | 88.4 | 83.8 |
| itbenchSre | 31.5 | 40.3 |
| MCP Atlas | 69 | 75.6 |
| MMLU-Pro | 86.4 | 86 |
| nl2repo | 54.2 | 42.7 |
| OmniScience Accuracy | 40.4 | 25.2 |
| OmniScience Non-Hallucination | 15.6 | 70.1 |
| scicode | 49.9 | 43.8 |
| simplebench | 46.3 | 55.1 |
| simpleqa_verified | 45 | 38.1 |
| SWE-bench Multilingual | 73.3 | 73.3 |
| SWE-bench Pro | 52.6 | 58.4 |
| TauBench V3 - Banking | 39.4 | 13.6 |
| Terminal-Bench 2.0 | 56.9 | 69 |
| Terminal-Bench 2.1 | 82.7 | 63.5 |
| Terminal-Bench Hard | 38.6 | 43.2 |
| toolathlon | 70.3 | 40.7 |
| τ²-Bench Telecom (AA run) | 95.6 | 97.7 |
| τ³-Bench | 22.9 | 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.