| Benchmark | Gemini 3.6 Flash | Qwen3 VL 4B (Reasoning) |
|---|---|---|
| AA Agentic Index | 40.5 | 14.4 |
| AA Intelligence | 52 | 7.7 |
| AA-LCR | 79 | 23 |
| AA-Omniscience | 22.1 | -68.4 |
| Artificial Analysis Coding Index | 69.2 | 6.7 |
| critpt | 10.6 | 0 |
| gdpval | 46.1 | 13.8 |
| GPQA Diamond | 92.8 | 64.1 |
| HLE | 40.8 | 4.6 |
| LVBench | 84.2 | 53.5 |
| MMMU-Pro | 83.2 | 57 |
| OmniScience Accuracy | 50 | 12.2 |
| OmniScience Non-Hallucination | 44.4 | 8.2 |
| OSWorld-Verified | 83 | 31.4 |
| scicode | 52.7 | 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.