Across 18 shared benchmarks, GPT-4o mini scores higher on 0 and Qwen3.5 122B A10B on 18. The widest gap is Terminal-Bench 2.1, where Qwen3.5 122B A10B scores 47.6 against 5.6. GPT-4o mini is the cheaper of the two on tracked API pricing ($0.15 against $0.40 per million input tokens).
| Benchmark | GPT-4o mini | Qwen3.5 122B A10B |
|---|---|---|
| AA Agentic Index | 1 | 21.3 |
| AA Intelligence | 7 | 32.8 |
| AI2D | 75.2 | 93.3 |
| Artificial Analysis Coding Index | 11.4 | 45.7 |
| gdpval | 0 | 24.3 |
| GPQA Diamond | 42.6 | 86.6 |
| GSM8K | 84.3 | 94.5 |
| HLE | 4.2 | 47.5 |
| IFBench | 31 | 76.1 |
| MathVista | 56.7 | 87.4 |
| MMLU-Pro | 61.7 | 86.7 |
| MMMU | 59.4 | 83.9 |
| MMMU-Pro | 41.5 | 76.9 |
| scicode | 22.9 | 42 |
| SWE-bench Verified | 8.7 | 72 |
| TauBench V3 - Banking | 2.9 | 15.3 |
| Terminal-Bench 2.1 | 5.6 | 47.6 |
| τ³-Bench | 2.9 | 13.6 |
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. Quoted rates are the price-setter row we currently track for each model — its direct or vendor-official listing where one exists (direct, direct), otherwise the lowest tracked offer.