Step 3.5 Flash vs Kimi K2 Thinking
Lado a lado. Cada métrica. Cada benchmark.
| Tipo | Step 3.5 Flash | Kimi K2 Thinking |
|---|---|---|
| Provider | ||
| puntuación promedio | 76.9 | 53.3 |
| Precio de entrada | $0.10 | $0.60 |
| Precio de salida | $0.30 | $2.50 |
| Ventana de contexto | 262K tokens (~131 books) | 262K tokens (~131 books) |
| Publicado el | 2026-01-29 | 2025-11-06 |
| Código abierto | Open Source | Open Source |
Puntuaciones de benchmark
6 benchmarks · Step 3.5 Flash: 5, Kimi K2 Thinking: 1
| Benchmark | Categoría | Step 3.5 Flash | Kimi K2 Thinking |
|---|---|---|---|
| OpenCompass — AIME2025 | math | 95.7 | 94.1 |
| OpenCompass — GPQA-Diamond | knowledge | 83.7 | 82.7 |
| OpenCompass — HLE | knowledge | 21.6 | 21.3 |
| OpenCompass — IFEval | language | 93.2 | 92.4 |
| OpenCompass — LiveCodeBenchV6 | coding | 83.9 | 77.1 |
| OpenCompass — MMLU-Pro | knowledge | 83.5 | 84.3 |