| Benchmark | Claude Opus 4.7 | Qwen3 VL 4B (Reasoning) |
|---|---|---|
| AA Agentic Index | 64.6 | 14.4 |
| AA Intelligence | 55 | 7.7 |
| AA-LCR | 75.3 | 23 |
| AA-Omniscience | 27.3 | -68.4 |
| Artificial Analysis Coding Index | 73.6 | 6.7 |
| critpt | 12 | 0 |
| gdpval | 58.6 | 13.8 |
| GPQA Diamond | 94.2 | 64.1 |
| HLE | 54.7 | 4.6 |
| IFBench | 58.6 | 36.6 |
| MMMU-Pro | 78.8 | 57 |
| OmniScience Accuracy | 48.9 | 12.2 |
| OmniScience Non-Hallucination | 57.7 | 8.2 |
| OSWorld-Verified | 82.8 | 31.4 |
| scicode | 54.5 | 17.1 |
| screenspot_pro_no_tools | 79.5 | 49.2 |
| Terminal-Bench Hard | 54.5 | 1.5 |
| τ²-Bench Telecom (AA run) | 88.6 | 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.