Qwen2-VL 2B Instruct, 4-bit MLX, for Lantern
A copy of mlx-community/Qwen2-VL-2B-Instruct-4bit, kept here for Lantern, an offline, private AI app for iPhone, iPad and Mac. The model runs entirely on the device: no account, no server, and nothing you type leaves it. The only network use is downloading these files once.
The weights are unchanged from the mlx-community conversion. The original model is Qwen/Qwen2-VL-2B-Instruct.
| Parameters | 2.2 billion |
| Precision | 4-bit, MLX format |
| Download | about 1.3 GB |
| Takes in | text and photos |
| Runs on | 6 GB iPhones and up, and any Apple silicon Mac. |
Good for: Reads a menu, a sign, a document or a diagram in a photo, and answers follow-up questions about it.
Use it in Lantern
Lantern already lists this model. To download this copy instead, open Settings, turn on Show advanced options, choose Add a model from Hugging Face and enter RabiatS/Qwen2-VL-2B-Instruct-4bit-Lantern. Lantern checks it against your device's memory before anything downloads.
Use it with mlx-lm on a Mac
pip install mlx-vlm
mlx_vlm.generate --model RabiatS/Qwen2-VL-2B-Instruct-4bit-Lantern --image photo.jpg --prompt "What is in this picture?"
Licence
Apache 2.0, as published with Qwen/Qwen2-VL-2B-Instruct.
Credit for the model goes to its authors, and for the MLX conversion to mlx-community.