DeepSeek V3 vs Gemini 3 Pro
Côte à côte. Chaque métrique. Chaque benchmark.
| Type | DeepSeek V3 | Gemini 3 Pro |
|---|---|---|
| Provider |  DeepSeek |  Google DeepMind |
| score moyen | 59.0 | 60.5 |
| Prix d'entrée | $0.32 | - |
| Prix de sortie | $0.89 | - |
| Fenêtre de contexte | 164K tokens (~82 books) | - |
| Sorti le | 2024-12-26 | 2024-01-01 |
| Code source ouvert | Open Source | Proprietary |
Scores de benchmark
11 benchmarks · DeepSeek V3: 0, Gemini 3 Pro: 11
| Benchmark | Catégorie | DeepSeek V3 | Gemini 3 Pro |
|---|---|---|---|
| Chatbot Arena Elo — Overall | arena | 1358.2 | 1486.2 |
| FrontierMath-2025-02-28-Private | math | 1.7 | 37.6 |
| GPQA diamond | knowledge | 42.0 | 90.2 |
| HELM — GPQA | knowledge | 53.8 | 80.3 |
| HELM — IFEval | language | 83.2 | 87.6 |
| HELM — MMLU-Pro | knowledge | 72.3 | 90.3 |
| HELM — Omni-MATH | math | 40.3 | 55.6 |
| HELM — WildBench | reasoning | 83.1 | 85.9 |
| OTIS Mock AIME 2024-2025 | math | 15.8 | 91.4 |
| SimpleBench | reasoning | 2.7 | 71.7 |
| WeirdML | coding | 36.1 | 69.9 |