Compare / Claude Opus 5 vs Gemini 3.5 Flash
Claude Opus 5 vs Gemini 3.5 Flash
21 results from 6 sources that measured both models. Each row is in the source's own units; there is no overall winner.
| Benchmark | Metric | Claude Opus 5 | Gemini 3.5 Flash |
|---|---|---|---|
| APEX-Agentsreported | Tasks passed% of tasks | 65.8% | 27.5% |
| Arena (formerly LMArena)reported | OverallArena rating | 1,490 | 1,461 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,488 | 1,476 |
| Arena (formerly LMArena)reported | Business, management and financeArena rating | 1,486 | 1,468 |
| Arena (formerly LMArena)reported | Creative writingArena rating | 1,472 | 1,466 |
| Arena (formerly LMArena)reported | Expert promptsArena rating | 1,543 | 1,491 |
| Arena (formerly LMArena)reported | Instruction followingArena rating | 1,497 | 1,465 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,491 | 1,477 |
| Arena (formerly LMArena)reported | Writing, literature and languageArena rating | 1,482 | 1,472 |
| Artificial Analysisreported | Artificial Analysis Coding Indexindex score | 78.0 | 70.1 |
| Artificial Analysisreported | Artificial Analysis Intelligence Indexindex score | 50.8 | 33.6 |
| Artificial Analysisreported | GPQA Diamond% of questions | 93.7% | 92.2% |
| Artificial Analysisreported | Humanity's Last Exam% of questions | 54.9% | 42.7% |
| Artificial Analysisreported | Long-context reasoning (AA-LCR)% of questions | 82.0% | 74.3% |
| Artificial Analysisreported | SciCode% of problems | 56.4% | 53.9% |
| Artificial Analysisreported | Terminal-Bench 2.1% of tasks | 89.1% | 78.7% |
| Artificial Analysisreported | Terminal-Bench 4.0% of tasks | 49.0% | 6.6% |
| Artificial Analysisreported | Τ-bench banking% of tasks | 44.7% | 32.2% |
| GDP.pdfreported | Professional document tasks% of rubric | 24.0% | 14.0% |
| SimpleQA Verified (Epoch AI)reported | Correct answers% of questions | 59.9% | 66.2% |
| Vending-Bench 2reported | Money after a yearUS dollars | $11,181.87 | $5,396.42 |
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.