| Benchmark | GPT-5.5 | Qwen3 VL 4B (Reasoning) |
|---|---|---|
| AA Agentic Index | 47.4 | 14.4 |
| AA Intelligence | 56.3 | 7.7 |
| AA-LCR | 79 | 23 |
| AA-Omniscience | 20.5 | -68.4 |
| Artificial Analysis Coding Index | 74.9 | 6.7 |
| critpt | 27.1 | 0 |
| gdpval | 49.5 | 13.8 |
| GPQA Diamond | 93.6 | 64.1 |
| HLE | 52.2 | 4.6 |
| IFBench | 75.9 | 36.6 |
| MMMU-Pro | 83.2 | 57 |
| OmniScience Accuracy | 58 | 12.2 |
| OmniScience Non-Hallucination | 12 | 8.2 |
| OSWorld-Verified | 78.7 | 31.4 |
| scicode | 56.1 | 17.1 |
| Terminal-Bench Hard | 60.6 | 1.5 |
| τ²-Bench Telecom (AA run) | 93.9 | 15.5 |
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.