Kimi K2 Thinking vs DeepSeek V3.2 Speciale
Lado a lado. Cada métrica. Cada benchmark.
| Tipo | Kimi K2 Thinking | DeepSeek V3.2 Speciale |
|---|---|---|
| Provider |  moonshotai |  DeepSeek |
| puntuación promedio | 53.3 | 78.2 |
| Precio de entrada | $0.60 | $0.40 |
| Precio de salida | $2.50 | $1.20 |
| Ventana de contexto | 262K tokens (~131 books) | 164K tokens (~82 books) |
| Publicado el | 2025-11-06 | 2025-12-01 |
| Código abierto | Open Source | Open Source |
Puntuaciones de benchmark
6 benchmarks · Kimi K2 Thinking: 1, DeepSeek V3.2 Speciale: 5
| Benchmark | Categoría | Kimi K2 Thinking | DeepSeek V3.2 Speciale |
|---|---|---|---|
| OpenCompass — AIME2025 | math | 94.1 | 96.0 |
| OpenCompass — GPQA-Diamond | knowledge | 82.7 | 86.7 |
| OpenCompass — HLE | knowledge | 21.3 | 28.6 |
| OpenCompass — IFEval | language | 92.4 | 91.7 |
| OpenCompass — LiveCodeBenchV6 | coding | 77.1 | 80.9 |
| OpenCompass — MMLU-Pro | knowledge | 84.3 | 85.5 |