| Benchmark | Opus 5 | Qwen3 VL 4B (Reasoning) |
|---|---|---|
| AA Agentic Index | 59.2 | 14.4 |
| AA Intelligence | 63.1 | 7.7 |
| AA-LCR | 78.7 | 23 |
| AA-Omniscience | 37.1 | -68.4 |
| Artificial Analysis Coding Index | 78 | 6.7 |
| critpt | 29.1 | 0 |
| gdpval | 67.2 | 13.8 |
| GPQA Diamond | 93.7 | 64.1 |
| HLE | 64.7 | 4.6 |
| MMMU-Pro | 84.7 | 57 |
| OmniScience Accuracy | 60.9 | 12.2 |
| OmniScience Non-Hallucination | 40.5 | 8.2 |
| OSWorld-Verified | 4 | 31.4 |
| scicode | 55.7 | 17.1 |
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.