Runs on Multi-GPU · Hugging Face ↗
AA coding index 61.8 (Artificial Analysis, frozen since 2026-09-11)
Moonshot's ~1T general MoE — openly released (unlike K3), and a notch above its own K2.7-Code sibling on the coding index.
Real GGUF file sizes = weight VRAM. Add the KV cache for your context (≈57 GB at 32K, fp16). Estimated from the original model config, not from the GGUF we serve. Less reliable than the rest of this page.
| Quantization | Size |
|---|---|
| Q2_K_XL | 340.5 GB |
| Q4_K_XL | 583.7 GB |
| Q8_K_XL | 594.5 GB |
| BF16 | 2053.2 GB |
~580 GB at Q4 — cluster-scale. Serve tensor / pipeline parallel with vLLM or SGLang. Open but not local for most; rent it per token unless you run a fleet.