Downloads · 30 days
10
16% of all-time downloads
Qwen3.5-4B-MLX-E3-3bit-g128
krisdcosta/Qwen3.5-4B-MLX-E3-3bit-g128
Qwen3.5-4B-MLX-E3-3bit-g128 is a machine learning model from krisdcosta. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
Mixed-precision MLX checkpoint for on-device VLM inference on Apple Silicon.
All-time downloads
63
Public
Parameters
794M
2.4 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors2.4 GB · 99%
Parameter types
How the weights are stored.
BF16400M · 50%
Model card
From the Hugging Face model README
Qwen3.5-4B-MLX-E3-3bit-g128
Mixed-precision MLX checkpoint for on-device VLM inference on Apple Silicon.
- Base model: Qwen3.5-4B
- Vision tower: BF16 (unquantized)
- LM backbone: 3-bit, group-size 128
- Framework: MLX (mlx-vlm 0.6.3)
- Conversion source: mlx-community/Qwen3.5-4B-MLX-bf16
- Mac inference peak: 4.499 GB
- Disk size: 2.2 GB
- Semantic validation: Pass on 3 benchmark images (simple scene, text-heavy dashboard, complex lab)
- Part of: Z Lab / UCSD on-device VLM deployment project
- Type
- qwen3_5
- Created
- Jul 10, 2026
- Updated
- Jul 10, 2026