Compare / Gemini 3.5 Flash vs GPT-5.4 Nano
Gemini 3.5 Flash vs GPT-5.4 Nano
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 Nano |
|---|---|---|---|
| BulletBenchmeasured by us | Cost per gameUS dollars | $0.0289 | $0.0050 |
| BulletBenchmeasured by us | Games lost on time% of games | 25.0% | 66.7% |
| BulletBenchmeasured by us | Invalid moves% of moves | 0.5% | 3.1% |
| BulletBenchmeasured by us | Ladder Eloladder Elo | 810 | 33 |
| BulletBenchmeasured by us | Median move timemilliseconds | 1.2 s | 1.1 s |
| BulletBenchmeasured by us | Cost per gameUS dollars | $0.0213 | $0.0025 |
| BulletBenchmeasured by us | Games lost on time% of games | 33.3% | 75.0% |
| BulletBenchmeasured by us | Invalid moves% of moves | 0.6% | 4.0% |
| BulletBenchmeasured by us | Ladder Eloladder Elo | 727 | 197 |
| BulletBenchmeasured by us | Median move timemilliseconds | 1.2 s | 1.0 s |
| Arena (formerly LMArena)reported | OverallArena rating | 1,476 | 1,418 |
| Arena (formerly LMArena)reported | Business, management and financeArena rating | 1,468 | 1,405 |
| Arena (formerly LMArena)reported | Creative writingArena rating | 1,466 | 1,335 |
| Arena (formerly LMArena)reported | Expert promptsArena rating | 1,491 | 1,438 |
| Arena (formerly LMArena)reported | Instruction followingArena rating | 1,465 | 1,387 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,477 | 1,401 |
| Arena (formerly LMArena)reported | Writing, literature and languageArena rating | 1,472 | 1,362 |
| Artificial Analysisreported | Artificial Analysis Coding Indexindex score | 70.1 | 56.1 |
| Artificial Analysisreported | Artificial Analysis Intelligence Indexindex score | 33.6 | 20.7 |
| Artificial Analysisreported | GPQA Diamond% of questions | 92.2% | 81.7% |
| Artificial Analysisreported | Humanity's Last Exam% of questions | 42.7% | 28.3% |
| Artificial Analysisreported | IFBench% of instructions | 76.3% | 75.9% |
| Artificial Analysisreported | Long-context reasoning (AA-LCR)% of questions | 74.3% | 76.7% |
| Artificial Analysisreported | SciCode% of problems | 53.9% | 47.2% |
| Artificial Analysisreported | Terminal-Bench 2.1% of tasks | 78.7% | 60.7% |
| Artificial Analysisreported | Terminal-Bench 4.0% of tasks | 6.6% | 0.5% |
| Artificial Analysisreported | Terminal-Bench Hard% of tasks | 46.2% | 42.4% |
| Artificial Analysisreported | Τ-bench banking% of tasks | 32.2% | 27.4% |
| Artificial Analysisreported | Τ²-bench telecom% of tasks | 95.6% | 76.0% |
| SimpleQA Verified (Epoch AI)reported | Correct answers% of questions | 66.2% | 11.7% |
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.