Gemini 2.5 Pro vs DeepSeek V3.2 Speciale
Lado a lado. Cada métrica. Cada benchmark.
| Tipo | Gemini 2.5 Pro | DeepSeek V3.2 Speciale |
|---|---|---|
| Provider |  Google DeepMind |  DeepSeek |
| pontuação média | 56.2 | 78.2 |
| Preço de entrada | $1.25 | $0.40 |
| Preço de saída | $10.00 | $1.20 |
| Janela de contexto | 1.0M tokens (~524 books) | 164K tokens (~82 books) |
| Lançado em | 2025-06-17 | 2025-12-01 |
| Código aberto | Proprietary | Open Source |
Pontuações de benchmark
9 benchmarks · Gemini 2.5 Pro: 3, DeepSeek V3.2 Speciale: 6
| Benchmark | Categoria | Gemini 2.5 Pro | DeepSeek V3.2 Speciale |
|---|---|---|---|
| Artificial Analysis — Agentic Index | speed | 32.7 | 0.0 |
| Artificial Analysis — Coding Index | speed | 31.9 | 37.9 |
| Artificial Analysis — Quality Index | speed | 34.6 | 29.4 |
| OpenCompass — AIME2025 | math | 88.7 | 96.0 |
| OpenCompass — GPQA-Diamond | knowledge | 84.7 | 86.7 |
| OpenCompass — HLE | knowledge | 21.1 | 28.6 |
| OpenCompass — IFEval | language | 90.0 | 91.7 |
| OpenCompass — LiveCodeBenchV6 | coding | 71.3 | 80.9 |
| OpenCompass — MMLU-Pro | knowledge | 85.8 | 85.5 |