Qwen3-30B-A3B-Thinking-2507

BaseRT .base builds of Qwen/Qwen3-30B-A3B-Thinking-2507 for fast local inference on Apple Silicon (Metal).

Files

File	Precision	Size
Qwen3-30B-A3B-Thinking-2507-Q4.base	4-bit	17.2 GB
Qwen3-30B-A3B-Thinking-2507-Q8.base	8-bit	31.5 GB

Usage

curl -LsSf https://basecompute.co/install.sh | sh
basert pull basecompute/Qwen3-30B-A3B-Thinking-2507
basert chat basecompute/Qwen3-30B-A3B-Thinking-2507

This is a reasoning model; responses include a thinking section.

The 8-bit build is best suited to systems with 32 GB or more of unified memory; the 4-bit build runs comfortably on 16–24 GB.

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