Moonshot AI · First MLX build 27 Jan 2026
Kimi-K2.5
MLX
Text modified-mit
A 1.0T parameter text generation model from Moonshot AI, with 3 MLX builds, from 3-bit to 8-bit.
- Parameters
- 1.0T From Hugging Face
- Quantisation
- 4-bit Kimi-K2.5
- Memory needed
- 667 GB Estimate, unified memory
- Download
- 594 GB Weights, estimated
- Downloads
- 122K This build, last 30 days
- Task
- Text generation Text
- MLX builds
- 3 On mlx-community
- Devices
- Mac Apple silicon
Memory
Will it fit?Needs about 667 GB at 4-bit
8 GB
16 GB
24 GB
32 GB
64 GB
128 GB
Fits with headroomTight, close other appsWon’t fit
Estimated from the size of the weights, plus a fifth for context and working memory and 2 GB for macOS and your apps. Long contexts need more.
Run it
Terminal · mlx-lm$ pip install -U mlx-lm
$ mlx_lm.generate --model mlx-community/Kimi-K2.5 --prompt "Hello"MLX builds
3 on mlx-community| Kimi-K2.5-3bit | 3-bit | 449 GB | 505 GB | 141 |
|---|---|---|---|---|
| Kimi-K2.5 | 4-bit | 594 GB | 667 GB | 122K |
| Kimi-K2.5-mlx-DQ3_K_M-q8 | 8-bit | 1091 GB | 1221 GB | 186 |
Sizes are estimated from Hugging Face’s parameter counts and the quantisation. Downloads are for the last 30 days.