8-bit Open on Hugging Face
Parameters
4.5B
Weights (est.)
≈ 5.1 GB
Mac memory (est.)
8 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-vlm on an Apple silicon Mac:
pip install mlx-vlmpython -m mlx_vlm.generate --model mlx-community/Kestrel-4B-Coder-8bit --image photo.jpg --prompt "Describe this image"
Details
- Base model
- unsloth/Qwen3.5-4B (finetune)
- Vendor
- unsloth
- Task
- Vision + text
- Quantisation
- 8-bit
- Licence
- apache-2.0
- Downloads
- 152 in the last 30 days · 152 all time · 0 likes
- Published
- 2026-11-02 · updated 2026-11-07 (4d ago)
- Languages
- en, zh, ko, ja, ru, es
- Tags
- qwen3_5, unsloth, merged, distillation, reasoning, code-generation, coder, tool-use, function-calling, agent, multi-token-prediction, speculative-decoding
Other MLX builds of unsloth/Qwen3.5-4B
| Build | Quantisation | Weights | Mac memory | Downloads |
|---|---|---|---|---|
| Kestrel-4B-Coder-4bit | 4-bit | ≈ 3.0 GB | 8 GB+ | 150 |
| Kestrel-4B-Coder-6bit | 6-bit | ≈ 4.1 GB | 8 GB+ | 68 |
| Kestrel-4B-Coder-8bit (this one) | 8-bit | ≈ 5.1 GB | 8 GB+ | 152 |