Compare / Claude Opus 4.5 vs Gemini 2.5 Flash
Claude Opus 4.5 vs Gemini 2.5 Flash
17 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 Opus 4.5 | Gemini 2.5 Flash |
|---|---|---|---|
| Arena (formerly LMArena)reported | OverallArena rating | 1,459 | 1,437 |
| Arena (formerly LMArena)reported | Business, management and financeArena rating | 1,472 | 1,399 |
| Arena (formerly LMArena)reported | Creative writingArena rating | 1,469 | 1,394 |
| Arena (formerly LMArena)reported | Expert promptsArena rating | 1,503 | 1,424 |
| Arena (formerly LMArena)reported | Instruction followingArena rating | 1,483 | 1,400 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,474 | 1,409 |
| Arena (formerly LMArena)reported | Writing, literature and languageArena rating | 1,467 | 1,402 |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Irrelevance detection% correct | 90.8% | 93.7% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Memory% correct | 73.8% | 41.3% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Multi-turn tasks% correct | 68.4% | 36.2% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Overall accuracy% correct | 77.5% | 56.2% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Relevance detection% correct | 68.8% | 75.0% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Single-turn calls (curated)% correct | 89.7% | 88.1% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Single-turn calls (user-contributed)% correct | 79.8% | 78.2% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Web search% correct | 84.5% | 62.0% |
| Vectara Hallucination Leaderboardreported | Answer rate% of documents | 98.7% | 99.0% |
| Vectara Hallucination Leaderboardreported | Hallucination rate% of summaries | 10.9% | 7.8% |
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.