Compare / Qwen3 235B A22B Thinking 2507 vs DeepSeek V3.2 Exp
Qwen3 235B A22B Thinking 2507 vs DeepSeek V3.2 Exp
17 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 | DeepSeek V3.2 Exp |
|---|---|---|---|
| Arena (formerly LMArena)reported | Business, management and financeArena rating | 1,404 | 1,422 |
| Arena (formerly LMArena)reported | Creative writingArena rating | 1,370 | 1,409 |
| Arena (formerly LMArena)reported | Expert promptsArena rating | 1,466 | 1,449 |
| Arena (formerly LMArena)reported | Instruction followingArena rating | 1,385 | 1,416 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,400 | 1,425 |
| Arena (formerly LMArena)reported | Writing, literature and languageArena rating | 1,377 | 1,410 |
| Artificial Analysisreported | Artificial Analysis Intelligence Indexindex score | 12.7 | 16.6 |
| Artificial Analysisreported | GPQA Diamond% of questions | 79.0% | 79.7% |
| Artificial Analysisreported | Humanity's Last Exam% of questions | 15.9% | 14.9% |
| Artificial Analysisreported | IFBench% of instructions | 51.2% | 54.1% |
| Artificial Analysisreported | Long-context reasoning (AA-LCR)% of questions | 72.0% | 72.3% |
| Artificial Analysisreported | Terminal-Bench Hard% of tasks | 13.6% | 31.1% |
| Artificial Analysisreported | Τ²-bench telecom% of tasks | 53.2% | 33.9% |
| UGI Leaderboardreported | Requested-length error% off the requested word count | 30.0% | 25.0% |
| UGI Leaderboardreported | Style adherencescore from 0 to 1 | 0.31 | 0.35 |
| UGI Leaderboardreported | Writing scorescore out of 100 | 32.4 | 54.0 |
| Vending-Bench 2reported | Money after a yearUS dollars | $-11.34 | $1,034.00 |
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.