Skip to content

Kimi-K2-Thinking-3bit

mlx-community/Kimi-K2-Thinking-3bit
3-bit Open on Hugging Face
Parameters
1.0T
Weights (est.)
≈ 449 GB
Mac memory (est.)
512 GB+

Size is estimated from the parameter count and quantisation. Memory allows a fifth extra for context plus 2 GB for the system; long contexts need more.

Run it

With mlx-lm on an Apple silicon Mac:

pip install mlx-lmmlx_lm.generate --model mlx-community/Kimi-K2-Thinking-3bit --prompt "Hello"

Details

Base model
moonshotai/Kimi-K2-Thinking (quantized)
Vendor
Moonshot · Kimi
Task
Text generation
Quantisation
3-bit
Licence
modified-mit
Downloads
51 in the last 30 days · 2.4K all time · 1 likes
Published
2025-11-25 · updated 2025-11-25 (11mo ago)
Tags
kimi_k2, custom_code

Other MLX builds of moonshotai/Kimi-K2-Thinking

Build Quantisation Weights Mac memory Downloads
Kimi-K2-Thinking-3bit (this one) 3-bit ≈ 449 GB 512 GB+ 51
Kimi-K2-Thinking-4bit 4-bit ≈ 578 GB > 512 GB 286
Kimi-K2-Thinking 4-bit ≈ 594 GB > 512 GB 1.6K