4-bit Open on Hugging Face
Parameters
121M
Weights (est.)
≈ 68 MB
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-lm on an Apple silicon Mac:
pip install mlx-lmmlx_lm.generate --model mlx-community/Qwen3.5-4B-MTP-4bit --prompt "Hello"
Details
- Base model
- Qwen/Qwen3.5-4B (quantized) · Qwen3.5 4B on the radar →
- Vendor
- Qwen
- Task
- Text generation
- Quantisation
- 4-bit
- Licence
- apache-2.0
- Downloads
- 594 in the last 30 days · 5.3K all time · 2 likes
- Published
- 2026-06-01 · updated 2026-06-01 (5mo ago)
- Languages
- en
- Tags
- qwen3_5_mtp, mlx-vlm, qwen3.5, qwen3.5-4b, qwen, speculative-decoding, draft-model
Other MLX builds of Qwen/Qwen3.5-4B
| Build | Quantisation | Weights | Mac memory | Downloads |
|---|---|---|---|---|
| Qwen3.5-4B-MTP-4bit (this one) | 4-bit | ≈ 68 MB | 8 GB+ | 594 |
| Qwen3.5-4B-MTP-5bit | 5-bit | ≈ 83 MB | 8 GB+ | 446 |
| Qwen3.5-4B-MTP-bf16 | bf16 | ≈ 241 MB | 8 GB+ | 201 |
| Ternary-Bonsai-2-4B-2bit | 2-bit | ≈ 2.0 GB | 8 GB+ | 547 |
| Qwen3.5-4B-OptiQ-4bit | 4-bit | ≈ 2.4 GB | 8 GB+ | 5.5K |
| Qwen3.5-4B-MLX-4bit | 4-bit | ≈ 3.0 GB | 8 GB+ | 102K |
| Qwen3.5-4B-MLX-8bit | 8-bit | ≈ 5.1 GB | 8 GB+ | 1.7K |
| Qwen3.5-4B-MLX-bf16 | bf16 | ≈ 9.1 GB | 16 GB+ | 357 |