DeepSeek · First MLX build 7 Sept 2024
DeepSeek-V2.5
MLX
deepseek
A 236B parameter model from DeepSeek, with one MLX build at 4-bit.
- Parameters
- 236B From Hugging Face
- Quantisation
- 4-bit DeepSeek-V2.5-MLX-AQ4_1_64
- Memory needed
- 151 GB Estimate, unified memory
- Download
- 133 GB Weights, estimated
- Downloads
- 52 This build, last 30 days
- Task
- —
- MLX builds
- 1 On mlx-community
- Devices
- Mac Apple silicon
Memory
Will it fit?Needs about 151 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 · Hugging Face CLI$ pip install -U huggingface_hub
$ hf download mlx-community/DeepSeek-V2.5-MLX-AQ4_1_64MLX builds
1 on mlx-community| DeepSeek-V2.5-MLX-AQ4_1_64 | 4-bit | 133 GB | 151 GB | 52 |
|---|
Sizes are estimated from Hugging Face’s parameter counts and the quantisation. Downloads are for the last 30 days.