Compare / Gemini 3.6 Flash vs Mistral Small 4
Gemini 3.6 Flash vs Mistral Small 4
22 results from 3 sources that measured both models. Each row is in the source's own units; there is no overall winner.
| Benchmark | Metric | Gemini 3.6 Flash | Mistral Small 4 |
|---|---|---|---|
| BulletBenchmeasured by us | Cost per gameUS dollars | $0.0113 | $0.0039 |
| BulletBenchmeasured by us | Games lost on time% of games | 41.7% | 25.0% |
| BulletBenchmeasured by us | Invalid moves% of moves | 0.3% | 0.6% |
| BulletBenchmeasured by us | Ladder Eloladder Elo | 628 | 367 |
| BulletBenchmeasured by us | Median move timemilliseconds | 1.6 s | 0.6 s |
| BulletBenchmeasured by us | Cost per gameUS dollars | $0.0050 | $0.0041 |
| BulletBenchmeasured by us | Games lost on time% of games | 91.7% | 0.0% |
| BulletBenchmeasured by us | Invalid moves% of moves | 0.0% | 1.0% |
| BulletBenchmeasured by us | Ladder Eloladder Elo | 33 | 571 |
| BulletBenchmeasured by us | Median move timemilliseconds | 1.5 s | 0.6 s |
| Artificial Analysisreported | Artificial Analysis Coding Indexindex score | 69.2 | 26.6 |
| Artificial Analysisreported | Artificial Analysis Intelligence Indexindex score | 34.0 | 11.3 |
| Artificial Analysisreported | GPQA Diamond% of questions | 92.8% | 76.9% |
| Artificial Analysisreported | Humanity's Last Exam% of questions | 40.8% | 9.9% |
| Artificial Analysisreported | Long-context reasoning (AA-LCR)% of questions | 80.0% | 49.7% |
| Artificial Analysisreported | SciCode% of problems | 53.4% | 38.8% |
| Artificial Analysisreported | Terminal-Bench 2.1% of tasks | 77.5% | 21.0% |
| Artificial Analysisreported | Terminal-Bench 4.0% of tasks | 7.1% | 0.0% |
| Artificial Analysisreported | Τ-bench banking% of tasks | 29.9% | 4.9% |
| UGI Leaderboardreported | Requested-length error% off the requested word count | 9.0% | 22.0% |
| UGI Leaderboardreported | Style adherencescore from 0 to 1 | 0.33 | 0.31 |
| UGI Leaderboardreported | Writing scorescore out of 100 | 69.8 | 40.3 |
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.