Alibaba Qwen · First MLX build 27 Nov 2024
QwQ-32B-Preview
MLX
Text apache-2.0
A 33B parameter text generation model from Alibaba Qwen, with 8 MLX builds, from 2-bit to bf16.
- Parameters
- 33B From Hugging Face
- Quantisation
- 6-bit QwQ-32B-Preview-6bit
- Memory needed
- 32 GB Estimate, unified memory
- Download
- 26.6 GB Weights, estimated
- Downloads
- 79 This build, last 30 days
- Task
- Text generation Text
- MLX builds
- 8 On mlx-community
- Devices
- Mac Apple silicon
Memory
Will it fit?Needs about 32 GB at 6-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/QwQ-32B-Preview-6bit --prompt "Hello"MLX builds
8 on mlx-community| QwQ-32B-Preview-2bit | 2-bit | 10.2 GB | 14 GB | 60 |
|---|---|---|---|---|
| QwQ-32B-Preview-3bit | 3-bit | 14.3 GB | 19 GB | 63 |
| QwQ-32B-Preview-4bit | 4-bit | 18.4 GB | 23 GB | 58 |
| Qwen_QwQ-32B-Preview_MLX-4bit | 4-bit | 18.4 GB | 23 GB | 56 |
| QwQ-32B-Preview-6bit | 6-bit | 26.6 GB | 32 GB | 79 |
| Qwen_QwQ-32B-Preview_MLX-8bit | 8-bit | 34.8 GB | 41 GB | 75 |
| QwQ-32B-Preview-8bit | 8-bit | 34.8 GB | 41 GB | 56 |
| QwQ-32B-Preview-bf16 | bf16 | 65.5 GB | 76 GB | 55 |
Sizes are estimated from Hugging Face’s parameter counts and the quantisation. Downloads are for the last 30 days.