Qwen3.5-2B

BaseRT .base builds of Qwen/Qwen3.5-2B for fast local inference on Apple Silicon (Metal).

Qwen3.5-2B is a hybrid-attention (Gated DeltaNet + periodic full attention) instruct model with a native vision tower. These bundles include the vision encoder — image inputs work out of the box.

Files

File	Precision	Size
Qwen3.5-2B-Q4.base	4-bit	1.3 GB
Qwen3.5-2B-Q8.base	8-bit	2.3 GB

Usage

curl -LsSf https://basecompute.co/install.sh | sh
basert pull basecompute/Qwen3.5-2B
basert chat basecompute/Qwen3.5-2B

Released under the apache-2.0 license, inherited from the base model.