Compare / Qwen3.5-122B-A10B vs GPT-5 Mini
Qwen3.5-122B-A10B vs GPT-5 Mini
21 results from 3 sources that measured both models. Each row is in the source's own units; there is no overall winner.
| Benchmark | Metric | Qwen3.5-122B-A10B | GPT-5 Mini |
|---|---|---|---|
| Arena (formerly LMArena)reported | OverallArena rating | 1,428 | 1,394 |
| Arena (formerly LMArena)reported | Business, management and financeArena rating | 1,422 | 1,381 |
| Arena (formerly LMArena)reported | Creative writingArena rating | 1,366 | 1,324 |
| Arena (formerly LMArena)reported | Expert promptsArena rating | 1,442 | 1,400 |
| Arena (formerly LMArena)reported | Instruction followingArena rating | 1,406 | 1,373 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,416 | 1,390 |
| Arena (formerly LMArena)reported | Writing, literature and languageArena rating | 1,387 | 1,352 |
| Artificial Analysisreported | Artificial Analysis Coding Indexindex score | 45.7 | 15.6 |
| Artificial Analysisreported | Artificial Analysis Intelligence Indexindex score | 17.7 | 20.6 |
| Artificial Analysisreported | GPQA Diamond% of questions | 85.7% | 82.8% |
| Artificial Analysisreported | Humanity's Last Exam% of questions | 25.2% | 21.5% |
| Artificial Analysisreported | IFBench% of instructions | 75.7% | 75.4% |
| Artificial Analysisreported | Long-context reasoning (AA-LCR)% of questions | 76.3% | 72.3% |
| Artificial Analysisreported | SciCode% of problems | 39.7% | 39.0% |
| Artificial Analysisreported | Terminal-Bench 2.1% of tasks | 47.6% | 3.7% |
| Artificial Analysisreported | Terminal-Bench 4.0% of tasks | 0.0% | 0.0% |
| Artificial Analysisreported | Terminal-Bench Hard% of tasks | 31.1% | 33.3% |
| Artificial Analysisreported | Τ-bench banking% of tasks | 15.3% | 15.5% |
| Artificial Analysisreported | Τ²-bench telecom% of tasks | 93.6% | 71.1% |
| Vectara Hallucination Leaderboardreported | Answer rate% of documents | 99.8% | 99.9% |
| Vectara Hallucination Leaderboardreported | Hallucination rate% of summaries | 11.2% | 12.9% |
Bold green marks the better value on that metric. Where a source reports ranges that overlap, the difference may not be meaningful; see the benchmark page for ranges.