Qwen3.6-35B-A3B

BaseRT .base builds of Qwen/Qwen3.6-35B-A3B for fast local inference on Apple Silicon (Metal).

A 35B mixture-of-experts reasoning model (3B active) with hybrid attention (Gated DeltaNet + periodic full attention). Q4 runs on 24 GB Apple Silicon; Q8 needs 48 GB+.

Files

File	Precision	Size
Qwen3.6-35B-A3B-Q4.base	4-bit	18 GB
Qwen3.6-35B-A3B-Q8.base	8-bit	33 GB

Usage

curl -LsSf https://basecompute.co/install.sh | sh
basert pull basecompute/Qwen3.6-35B-A3B
basert chat basecompute/Qwen3.6-35B-A3B

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