| Benchmark | GPT-5.4 | Qwen3.7 Plus Preview |
|---|---|---|
| AA Agentic Index | 58.2 | 20.8 |
| AA Intelligence | 53.1 | 39.4 |
| AA-LCR | 77.7 | 69 |
| AA-Omniscience | 5.8 | 2.4 |
| Apex (Pass@1) | 54.1 | 22.7 |
| apexAgents | 33.3 | 22.4 |
| arena_elo | 1477 | 1458 |
| arena_text_factuality | 1476 | 1454 |
| arena_vision | 1277 | 1263 |
| Artificial Analysis Coding Index | 71.1 | 55.9 |
| baby_vision | 49.7 | 70.4 |
| charxiv_rq | 82.8 | 85.9 |
| critpt | 23.4 | 9.1 |
| DeepSWE 1.1 | 52 | 14.2 |
| gdpval | 50.1 | 22.2 |
| GPQA Diamond | 93 | 90.3 |
| HLE | 43.7 | 35.6 |
| hmmt_feb_2026 | 97.7 | 92.9 |
| IFBench | 73.9 | 79.1 |
| imo_answer_bench | 91.4 | 86 |
| mathvision | 92 | 90.3 |
| MCP Atlas | 70.6 | 73.2 |
| mcpmark | 62.5 | 58.7 |
| MMLU-Pro | 87.5 | 88.5 |
| MMMU-Pro | 81.2 | 80.5 |
| nl2repo | 41.3 | 41.1 |
| OmniDocBench 1.5 | 89.1 | 91.4 |
| OmniScience Accuracy | 50.9 | 22.5 |
| OmniScience Non-Hallucination | 17.4 | 74.5 |
| OSWorld-Verified | 75 | 73.3 |
| scicode | 56.6 | 51.3 |
| SWE-bench Multilingual | 71.7 | 75.8 |
| SWE-bench Pro | 59.1 | 57.6 |
| TauBench V3 - Banking | 39.6 | 17.9 |
| Terminal-Bench 2.0 | 75.1 | 70.3 |
| Terminal-Bench 2.1 | 78.3 | 61 |
| Terminal-Bench Hard | 57.6 | 47 |
| τ²-Bench Telecom (AA run) | 87.1 | 93 |
| τ³-Bench | 72.9 | 17.9 |
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.