Alibaba Qwen · First MLX build 14 Aug 2026 · In Core AI since 16 Oct 2026
Qwen3.8-27B
MLXCore AI
TextVision apache-2.0 Trending
A 28B parameter vision and text model from Alibaba Qwen, with 14 MLX builds, from 4-bit to bf16 and Apple Core AI export recipes for macOS.
- Parameters
- 28B From Hugging Face
- Quantisation
- 4-bit Qwen3.8-27B-4bit
- Memory needed
- 20 GB Estimate, unified memory
- Download
- 16.1 GB Weights, estimated
- Downloads
- 110K This build, last 30 days
- Task
- Vision and text Text, Vision
- MLX builds
- 14 On mlx-community
- Devices
- Mac Apple silicon
- Parameters
- 28B From Hugging Face
- Compression
- None Apple’s preset
- Memory needed
- — Parameter count not published
- Context
- 256K tokens Fixed at export
- Precision
- BF16 Compute precision
- Platforms
- macOS Mac
- Export presets
- 1 In Apple’s registry
- Listed by Apple
- 16 Oct 2026 First registry commit
Memory
Will it fit?Needs about 20 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-vlm$ pip install -U mlx-vlm
$ python -m mlx_vlm.generate --model mlx-community/Qwen3.8-27B-4bit --image photo.jpg --prompt "Describe this image"Memory
There is no memory estimate: the parameter count for this model is not published on Hugging Face.
Export it
Terminal · uv$ git clone https://github.com/apple/coreai-models
$ cd coreai-models
$ uv run coreai.llm.export qwen3.8-27b
Apple publishes the recipe, not a ready-made package: these commands convert the original weights to a Core AI .aimodel on your Mac. See the repository for requirements.
MLX builds
14 on mlx-community| Qwen3.8-27B-MTP-4bit | 4-bit | 239 MB | 3 GB | 7.3K |
|---|---|---|---|---|
| Qwen3.8-27B-MTP-nvfp4 | 4-bit | 239 MB | 3 GB | 420 |
| Qwen3.8-27B-MTP-mxfp4 | MXFP4 | 239 MB | 3 GB | 316 |
| Qwen3.8-27B-MTP-8bit | 8-bit | 451 MB | 3 GB | 2.4K |
| Qwen3.8-27B-MTP-mxfp8 | MXFP8 | 451 MB | 3 GB | 474 |
| Qwen3.8-27B-MTP-bf16 | bf16 | 849 MB | 3 GB | 863 |
| Qwen3.8-27B-4bit | 4-bit | 16.1 GB | 20 GB | 110K |
| Qwen3.8-27B-OptiQ-4bit | 4-bit | 16.1 GB | 20 GB | 8.5K |
| Qwen3.8-27B-mxfp4 | MXFP4 | 16.1 GB | 20 GB | 2.7K |
| Qwen3.8-27B-nvfp4 | 4-bit | 16.1 GB | 20 GB | 1.1K |
| Qwen3.8-27B-GPTQ-4bit | 4-bit | 16.1 GB | 20 GB | 909 |
| Qwen3.8-27B-8bit | 8-bit | 29.5 GB | 35 GB | 27K |
| Qwen3.8-27B-mxfp8 | MXFP8 | 29.5 GB | 35 GB | 697 |
| Qwen3.8-27B-bf16 | bf16 | 54.7 GB | 64 GB | 6.0K |
Sizes are estimated from Hugging Face’s parameter counts and the quantisation. Downloads are for the last 30 days.
Core AI export presets
From Apple’s registry| Preset | Platform | Compression | Precision | Context |
|---|---|---|---|---|
| qwen3.8-27b | macOS | None | BF16 | 256K |