| Benchmark | Gemini 1.5 Pro | Qwen3.5 35B A3B |
|---|---|---|
| AA Intelligence | 10 | 29.9 |
| AI2D | 80.9 | 92.6 |
| Artificial Analysis Coding Index | 23.6 | 37 |
| GPQA Diamond | 59.1 | 84.5 |
| GSM8K | 95.2 | 90.1 |
| hellaswag | 93.3 | 85.6 |
| HLE | 4.9 | 47.4 |
| humaneval | 86.6 | 66.5 |
| ifeval | 89.4 | 91.9 |
| MathVista | 70.6 | 86.2 |
| mmlu | 86.8 | 81.1 |
| MMLU-Pro | 75.8 | 85.3 |
| MMMU | 68.4 | 81.4 |
| MMMU-Pro | 55 | 75.1 |
| OCRBench | 800 | 91 |
| ocrbench_v2 | 51.6 | 65.3 |
| OpenBookQA | 95.2 | 44.2 |
| RULER 1M | 85 | 56.4 |
| RULER 256K | 91.6 | 82.4 |
| scicode | 29.5 | 37.7 |
| SWE-bench Verified | 34.2 | 70 |
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.