Compare / Gemini 3.5 Flash vs GPT-5.4 Mini
Gemini 3.5 Flash vs GPT-5.4 Mini
30 results from 4 sources that measured both models. Each row is in the source's own units; there is no overall winner.
| Benchmark | Metric | Gemini 3.5 Flash | GPT-5.4 Mini |
|---|---|---|---|
| BulletBenchmeasured by us | Cost per gameUS dollars | $0.0289 | $0.0205 |
| BulletBenchmeasured by us | Games lost on time% of games | 25.0% | 50.0% |
| BulletBenchmeasured by us | Invalid moves% of moves | 0.5% | 0.0% |
| BulletBenchmeasured by us | Ladder Eloladder Elo | 810 | 197 |
| BulletBenchmeasured by us | Median move timemilliseconds | 1.2 s | 0.9 s |
| BulletBenchmeasured by us | Cost per gameUS dollars | $0.0213 | $0.0136 |
| BulletBenchmeasured by us | Games lost on time% of games | 33.3% | 0.0% |
| BulletBenchmeasured by us | Invalid moves% of moves | 0.6% | 0.0% |
| BulletBenchmeasured by us | Ladder Eloladder Elo | 727 | 571 |
| BulletBenchmeasured by us | Median move timemilliseconds | 1.2 s | 0.9 s |
| Arena (formerly LMArena)reported | OverallArena rating | 1,476 | 1,447 |
| Arena (formerly LMArena)reported | Business, management and financeArena rating | 1,468 | 1,460 |
| Arena (formerly LMArena)reported | Creative writingArena rating | 1,466 | 1,402 |
| Arena (formerly LMArena)reported | Expert promptsArena rating | 1,491 | 1,480 |
| Arena (formerly LMArena)reported | Instruction followingArena rating | 1,465 | 1,434 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,477 | 1,447 |
| Arena (formerly LMArena)reported | Writing, literature and languageArena rating | 1,472 | 1,424 |
| Artificial Analysisreported | Artificial Analysis Coding Indexindex score | 70.1 | 56.1 |
| Artificial Analysisreported | Artificial Analysis Intelligence Indexindex score | 33.6 | 24.1 |
| Artificial Analysisreported | GPQA Diamond% of questions | 92.2% | 87.5% |
| Artificial Analysisreported | Humanity's Last Exam% of questions | 42.7% | 28.1% |
| Artificial Analysisreported | IFBench% of instructions | 76.3% | 73.3% |
| Artificial Analysisreported | Long-context reasoning (AA-LCR)% of questions | 74.3% | 77.0% |
| Artificial Analysisreported | SciCode% of problems | 53.9% | 52.1% |
| Artificial Analysisreported | Terminal-Bench 2.1% of tasks | 78.7% | 59.2% |
| Artificial Analysisreported | Terminal-Bench 4.0% of tasks | 6.6% | 2.0% |
| Artificial Analysisreported | Terminal-Bench Hard% of tasks | 46.2% | 52.3% |
| Artificial Analysisreported | Τ-bench banking% of tasks | 32.2% | 25.6% |
| Artificial Analysisreported | Τ²-bench telecom% of tasks | 95.6% | 83.3% |
| SimpleQA Verified (Epoch AI)reported | Correct answers% of questions | 66.2% | 29.4% |
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.