Compare / Mistral Medium 3.5 vs GPT-5.4 Nano
Mistral Medium 3.5 vs GPT-5.4 Nano
29 results from 3 sources that measured both models. Each row is in the source's own units; there is no overall winner.
| Benchmark | Metric | Mistral Medium 3.5 | GPT-5.4 Nano |
|---|---|---|---|
| BulletBenchmeasured by us | Cost per gameUS dollars | $0.0354 | $0.0050 |
| BulletBenchmeasured by us | Games lost on time% of games | 0.0% | 66.7% |
| BulletBenchmeasured by us | Invalid moves% of moves | 0.4% | 3.1% |
| BulletBenchmeasured by us | Ladder Eloladder Elo | 572 | 33 |
| BulletBenchmeasured by us | Median move timemilliseconds | 0.5 s | 1.1 s |
| BulletBenchmeasured by us | Cost per gameUS dollars | $0.0297 | $0.0025 |
| BulletBenchmeasured by us | Games lost on time% of games | 0.0% | 75.0% |
| BulletBenchmeasured by us | Invalid moves% of moves | 0.0% | 4.0% |
| BulletBenchmeasured by us | Ladder Eloladder Elo | 529 | 197 |
| BulletBenchmeasured by us | Median move timemilliseconds | 0.5 s | 1.0 s |
| Arena (formerly LMArena)reported | OverallArena rating | 1,430 | 1,418 |
| Arena (formerly LMArena)reported | Business, management and financeArena rating | 1,430 | 1,405 |
| Arena (formerly LMArena)reported | Creative writingArena rating | 1,398 | 1,335 |
| Arena (formerly LMArena)reported | Expert promptsArena rating | 1,434 | 1,438 |
| Arena (formerly LMArena)reported | Instruction followingArena rating | 1,421 | 1,387 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,426 | 1,401 |
| Arena (formerly LMArena)reported | Writing, literature and languageArena rating | 1,404 | 1,362 |
| Artificial Analysisreported | Artificial Analysis Coding Indexindex score | 46.9 | 56.1 |
| Artificial Analysisreported | Artificial Analysis Intelligence Indexindex score | 14.2 | 20.7 |
| Artificial Analysisreported | GPQA Diamond% of questions | 74.8% | 81.7% |
| Artificial Analysisreported | Humanity's Last Exam% of questions | 13.8% | 28.3% |
| Artificial Analysisreported | IFBench% of instructions | 68.8% | 75.9% |
| Artificial Analysisreported | Long-context reasoning (AA-LCR)% of questions | 69.3% | 76.7% |
| Artificial Analysisreported | SciCode% of problems | 40.2% | 47.2% |
| Artificial Analysisreported | Terminal-Bench 2.1% of tasks | 50.6% | 60.7% |
| Artificial Analysisreported | Terminal-Bench 4.0% of tasks | 0.0% | 0.5% |
| Artificial Analysisreported | Terminal-Bench Hard% of tasks | 33.3% | 42.4% |
| Artificial Analysisreported | Τ-bench banking% of tasks | 15.1% | 27.4% |
| Artificial Analysisreported | Τ²-bench telecom% of tasks | 94.2% | 76.0% |
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.