| Benchmark | GPT-4o mini | Qwen3.5 35B A3B |
|---|---|---|
| AA Agentic Index | 1 | 44.1 |
| AA Intelligence | 7 | 29.9 |
| AI2D | 75.2 | 92.6 |
| Artificial Analysis Coding Index | 11.4 | 37 |
| gdpval | 0 | 20.3 |
| GPQA Diamond | 42.6 | 84.5 |
| GSM8K | 84.3 | 90.1 |
| HLE | 4.2 | 47.4 |
| humaneval | 87.2 | 66.5 |
| IFBench | 31 | 72.5 |
| MathVista | 56.7 | 86.2 |
| mmlu | 81.8 | 81.1 |
| MMMU | 59.4 | 81.4 |
| MMMU-Pro | 41.5 | 75.1 |
| OpenBookQA | 92 | 44.2 |
| scicode | 22.9 | 37.7 |
| SWE-bench Verified | 8.7 | 70 |
| TauBench V3 - Banking | 2.9 | 4.9 |
| Terminal-Bench 2.1 | 5.6 | 40.8 |
| τ³-Bench | 2.9 | 68.9 |
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.