Compare / Qwen3 235B A22B Instruct 2507 vs GPT-4.1
Qwen3 235B A22B Instruct 2507 vs GPT-4.1
18 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 235B A22B Instruct 2507 | GPT-4.1 |
|---|---|---|---|
| Arena (formerly LMArena)reported | OverallArena rating | 1,431 | 1,408 |
| Arena (formerly LMArena)reported | Business, management and financeArena rating | 1,432 | 1,415 |
| Arena (formerly LMArena)reported | Creative writingArena rating | 1,378 | 1,402 |
| Arena (formerly LMArena)reported | Expert promptsArena rating | 1,448 | 1,405 |
| Arena (formerly LMArena)reported | Instruction followingArena rating | 1,414 | 1,403 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,422 | 1,415 |
| Arena (formerly LMArena)reported | Writing, literature and languageArena rating | 1,396 | 1,405 |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Irrelevance detection% correct | 81.7% | 86.5% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Memory% correct | 23.9% | 23.9% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Multi-turn tasks% correct | 45.4% | 38.9% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Overall accuracy% correct | 52.1% | 54.0% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Relevance detection% correct | 93.8% | 100.0% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Single-turn calls (curated)% correct | 90.3% | 88.7% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Single-turn calls (user-contributed)% correct | 78.7% | 78.9% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Web search% correct | 54.0% | 68.0% |
| UGI Leaderboardreported | Requested-length error% off the requested word count | 15.0% | 6.0% |
| UGI Leaderboardreported | Style adherencescore from 0 to 1 | 0.36 | 0.33 |
| UGI Leaderboardreported | Writing scorescore out of 100 | 49.0 | 67.5 |
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.