Compare / Gemma 3 12B vs Mistral Medium 3
Gemma 3 12B vs Mistral Medium 3
17 results from 3 sources that measured both models. Each row is in the source's own units; there is no overall winner.
| Benchmark | Metric | Gemma 3 12B | Mistral Medium 3 |
|---|---|---|---|
| Arena (formerly LMArena)reported | Business, management and financeArena rating | 1,386 | 1,387 |
| Arena (formerly LMArena)reported | Creative writingArena rating | 1,332 | 1,364 |
| Arena (formerly LMArena)reported | Expert promptsArena rating | 1,274 | 1,379 |
| Arena (formerly LMArena)reported | Instruction followingArena rating | 1,321 | 1,366 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,342 | 1,387 |
| Arena (formerly LMArena)reported | Writing, literature and languageArena rating | 1,313 | 1,370 |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Irrelevance detection% correct | 70.3% | 92.0% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Memory% correct | 27.5% | 23.0% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Multi-turn tasks% correct | 5.8% | 10.8% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Overall accuracy% correct | 30.4% | 37.7% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Relevance detection% correct | 93.8% | 75.0% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Single-turn calls (curated)% correct | 79.4% | 85.3% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Single-turn calls (user-contributed)% correct | 74.2% | 68.0% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Web search% correct | 4.0% | 39.0% |
| UGI Leaderboardreported | Requested-length error% off the requested word count | 9.0% | 20.0% |
| UGI Leaderboardreported | Style adherencescore from 0 to 1 | 0.29 | 0.35 |
| UGI Leaderboardreported | Writing scorescore out of 100 | 29.9 | 32.6 |
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.