Compare / Qwen3 235B A22B Thinking 2507 vs GPT-5 Mini
Qwen3 235B A22B Thinking 2507 vs GPT-5 Mini
20 results from 4 sources that measured both models. Each row is in the source's own units; there is no overall winner.
| Benchmark | Metric | Qwen3 235B A22B Thinking 2507 | GPT-5 Mini |
|---|---|---|---|
| Arena (formerly LMArena)reported | Business, management and financeArena rating | 1,404 | 1,381 |
| Arena (formerly LMArena)reported | Creative writingArena rating | 1,370 | 1,324 |
| Arena (formerly LMArena)reported | Expert promptsArena rating | 1,466 | 1,400 |
| Arena (formerly LMArena)reported | Instruction followingArena rating | 1,385 | 1,373 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,400 | 1,390 |
| Arena (formerly LMArena)reported | Writing, literature and languageArena rating | 1,377 | 1,352 |
| Artificial Analysisreported | Artificial Analysis Coding Indexindex score | 22.1 | 15.6 |
| Artificial Analysisreported | Artificial Analysis Intelligence Indexindex score | 12.7 | 20.6 |
| Artificial Analysisreported | GPQA Diamond% of questions | 79.0% | 82.8% |
| Artificial Analysisreported | Humanity's Last Exam% of questions | 15.9% | 21.5% |
| Artificial Analysisreported | IFBench% of instructions | 51.2% | 75.4% |
| Artificial Analysisreported | Long-context reasoning (AA-LCR)% of questions | 72.0% | 72.3% |
| Artificial Analysisreported | SciCode% of problems | 41.4% | 39.0% |
| Artificial Analysisreported | Terminal-Bench 2.1% of tasks | 12.0% | 3.7% |
| Artificial Analysisreported | Terminal-Bench 4.0% of tasks | 0.0% | 0.0% |
| Artificial Analysisreported | Terminal-Bench Hard% of tasks | 13.6% | 33.3% |
| Artificial Analysisreported | Τ-bench banking% of tasks | 7.8% | 15.5% |
| Artificial Analysisreported | Τ²-bench telecom% of tasks | 53.2% | 71.1% |
| SimpleQA Verified (Epoch AI)reported | Correct answers% of questions | 40.4% | 21.6% |
| Vending-Bench 2reported | Money after a yearUS dollars | $-11.34 | $-31.18 |
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.