| Benchmark | Gemini 3.1 Pro | Qwen3 VL 4B (Reasoning) |
|---|---|---|
| AA Agentic Index | 23 | 14.4 |
| AA Intelligence | 47.7 | 7.7 |
| AA-LCR | 79 | 23 |
| AA-Omniscience | 32.9 | -68.4 |
| Artificial Analysis Coding Index | 68.8 | 6.7 |
| critpt | 17.7 | 0 |
| gdpval | 23.3 | 13.8 |
| GPQA Diamond | 94.3 | 64.1 |
| HLE | 51.4 | 4.6 |
| IFBench | 77.1 | 36.6 |
| LiveCodeBench | 91.7 | 51.3 |
| LiveCodeBench v6 | 91.7 | 51.3 |
| mathvision | 89.8 | 60 |
| MMLU-Pro | 91 | 73.6 |
| MMMU-Pro | 83 | 57 |
| OmniScience Accuracy | 55.3 | 12.2 |
| OmniScience Non-Hallucination | 50.1 | 8.2 |
| OSWorld-Verified | 76.2 | 31.4 |
| scicode | 59 | 17.1 |
| Terminal-Bench Hard | 68.5 | 1.5 |
| τ²-Bench Telecom (AA run) | 99.3 | 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.