Moonshot AI · 1.06T (32B active) · 専家混合（MoE）
Natively multimodal 1T MoE — 32B active, frontier agentic 看看你的 GPU 或 Mac 跑不跑得動 Kimi K2.6——最低 591.2 GB，建議 985.3 GB。
2026-04256K context
専家混合（MoE）
専家総数： 384
启用専家： 8
启用参数： 32.0B
| 量化 | 位元 | VRAM | 品質 | 状態 |
|---|---|---|---|---|
| Q2_K | 2 | 339.2 GB | low | — |
| Q3_K_M | 3 | 474.7 GB | moderate | — |
| Q4_K_M | 4 | 542.4 GB | good | — |
| Q5_K_M | 5 | 677.9 GB | good | — |
| Q6_K | 6 | 813.4 GB | excellent | — |
| Q8_0 | 8 | 1084.4 GB | excellent | — |
| F16 | 16 | 2168.2 GB | lossless | — |
Can I run Kimi K2.6 locally?
- Can I run Kimi K2.6 locally?
- Kimi K2.6 needs about 591.2 GB of memory at a minimum and 985.3 GB recommended. Open this page to grade it against your GPU or Mac, then run it with runai, Ollama or LM Studio.
- How much VRAM does Kimi K2.6 need?
- At Q4_K_M, Kimi K2.6 uses about 542.4 GB of VRAM. Higher quants need more memory; lower quants fit tighter cards with a quality tradeoff.