Compare / Claude Haiku 4.5 vs Gemini 3.6 Flash
Claude Haiku 4.5 vs Gemini 3.6 Flash
21 results from 3 sources that measured both models. Each row is in the source's own units; there is no overall winner.
| Benchmark | Metric | Claude Haiku 4.5 | Gemini 3.6 Flash |
|---|---|---|---|
| BulletBenchmeasured by us | Cost per gameUS dollars | $0.0163 | $0.0113 |
| BulletBenchmeasured by us | Games lost on time% of games | 25.0% | 41.7% |
| BulletBenchmeasured by us | Invalid moves% of moves | 0.3% | 0.3% |
| BulletBenchmeasured by us | Ladder Eloladder Elo | 445 | 628 |
| BulletBenchmeasured by us | Median move timemilliseconds | 1.2 s | 1.6 s |
| BulletBenchmeasured by us | Cost per gameUS dollars | $0.0095 | $0.0050 |
| BulletBenchmeasured by us | Games lost on time% of games | 83.3% | 91.7% |
| BulletBenchmeasured by us | Invalid moves% of moves | 0.0% | 0.0% |
| BulletBenchmeasured by us | Ladder Eloladder Elo | 33 | 33 |
| BulletBenchmeasured by us | Median move timemilliseconds | 1.3 s | 1.5 s |
| Arena (formerly LMArena)reported | OverallArena rating | 1,419 | 1,453 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,418 | 1,470 |
| Arena (formerly LMArena)reported | Business, management and financeArena rating | 1,420 | 1,472 |
| Arena (formerly LMArena)reported | Creative writingArena rating | 1,387 | 1,469 |
| Arena (formerly LMArena)reported | Expert promptsArena rating | 1,452 | 1,500 |
| Arena (formerly LMArena)reported | Instruction followingArena rating | 1,415 | 1,473 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,414 | 1,482 |
| Arena (formerly LMArena)reported | Writing, literature and languageArena rating | 1,399 | 1,472 |
| UGI Leaderboardreported | Requested-length error% off the requested word count | 6.0% | 9.0% |
| UGI Leaderboardreported | Style adherencescore from 0 to 1 | 0.39 | 0.33 |
| UGI Leaderboardreported | Writing scorescore out of 100 | 52.2 | 69.8 |
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.