| Benchmark | GPT-5.4 | Qwen3.6 27B |
|---|---|---|
| AA Agentic Index | 58.2 | 27.5 |
| AA Intelligence | 53.1 | 37.7 |
| AA-LCR | 77.7 | 73.3 |
| AA-Omniscience | 5.8 | -20 |
| aime_2026 | 99.2 | 94.1 |
| Artificial Analysis Coding Index | 71.1 | 53.7 |
| baby_vision | 49.7 | 28.9 |
| charxiv_rq | 82.8 | 78.4 |
| critpt | 23.4 | 1.1 |
| DeepSWE 1.1 | 52 | 13.3 |
| gdpval | 50.1 | 32 |
| GPQA Diamond | 93 | 87.8 |
| HLE | 43.7 | 24 |
| hmmt_feb_2026 | 97.7 | 84.3 |
| hmmt_nov_2025 | 95.8 | 90.7 |
| IFBench | 73.9 | 69.1 |
| imo_answer_bench | 91.4 | 80.8 |
| livebench_agentic_coding | 53.8 | 39.3 |
| livebench_coding | 77.5 | 71.8 |
| livebench_data_analysis | 79.3 | 70.4 |
| livebench_instruction_following | 70.2 | 53.2 |
| livebench_language | 82.6 | 63.3 |
| livebench_math | 94.2 | 79.9 |
| livebench_reasoning | 88.1 | 70.3 |
| mathvision | 92 | 85.1 |
| MMLU-Pro | 87.5 | 86.2 |
| MMMU-Pro | 81.2 | 75.8 |
| nl2repo | 41.3 | 36.2 |
| OmniDocBench 1.5 | 89.1 | 89.4 |
| OmniScience Accuracy | 50.9 | 19.6 |
| OmniScience Non-Hallucination | 17.4 | 50.7 |
| OSWorld-Verified | 75 | 63.9 |
| scicode | 56.6 | 39.8 |
| SWE-bench Multilingual | 71.7 | 71.3 |
| SWE-bench Pro | 59.1 | 53.5 |
| TauBench V3 - Banking | 39.6 | 16.7 |
| Terminal-Bench 2.0 | 75.1 | 59.3 |
| Terminal-Bench 2.1 | 78.3 | 60.7 |
| Terminal-Bench Hard | 57.6 | 34.8 |
| τ²-Bench Telecom (AA run) | 87.1 | 94.2 |
| τ³-Bench | 72.9 | 15.3 |
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.