Gemini 3 Flash Preview vs Claude Opus 4.5
Lado a lado. Cada métrica. Cada benchmark.
| Tipo | Gemini 3 Flash Preview | Claude Opus 4.5 |
|---|---|---|
| Provider |  Google DeepMind |  Anthropic |
| pontuação média | 49.1 | 45.4 |
| Preço de entrada | $0.50 | $5.00 |
| Preço de saída | $3.00 | $25.00 |
| Janela de contexto | 1.0M tokens (~524 books) | 200K tokens (~100 books) |
| Lançado em | 2025-12-17 | 2025-11-24 |
| Código aberto | Proprietary | Proprietary |
Pontuações de benchmark
20 benchmarks · Gemini 3 Flash Preview: 9, Claude Opus 4.5: 10
| Benchmark | Categoria | Gemini 3 Flash Preview | Claude Opus 4.5 |
|---|---|---|---|
| APEX-Agents | agentic | 24.0 | 18.4 |
| ARC-AGI | reasoning | 21.5 | 80.0 |
| ARC-AGI-2 | reasoning | 33.6 | 37.6 |
| Chatbot Arena Elo — Coding | arena | 1436.4 | 1465.2 |
| Chatbot Arena Elo — Overall | arena | 1473.9 | 1467.7 |
| Chess Puzzles | knowledge | 38.0 | 12.0 |
| FrontierMath-2025-02-28-Private | math | 35.6 | 20.7 |
| FrontierMath-Tier-4-2025-07-01-Private | math | 4.2 | 4.2 |
| GeoBench | knowledge | 88.0 | 75.0 |
| GPQA diamond | knowledge | 77.6 | 81.4 |
| GSO-Bench | coding | 9.8 | 26.5 |
| MCP Atlas | agentic | 57.4 | 62.3 |
| OTIS Mock AIME 2024-2025 | math | 92.8 | 86.1 |
| SciPredict | knowledge | 22.2 | 23.1 |
| SimpleBench | reasoning | 53.3 | 54.4 |
| SimpleQA Verified | knowledge | 67.4 | 41.8 |
| SWE-Bench verified | coding | 75.4 | 76.7 |
| Terminal Bench | coding | 64.3 | 63.1 |
| VPCT | knowledge | 58.9 | 10.0 |
| WeirdML | coding | 61.6 | 63.7 |