MiaAI-Lab

Qwen3.8 27B for 16 GB GPUs

The same Qwen3.8-27B family sized for a 16–32 GB NVIDIA card instead of a Spark. Mia picks the quant for the VRAM you have. This is the recipe for a desktop with a single consumer card, not a cluster.

Author: MiaAI-Lab
Last commit: 2026-09-10 · from the author's repository
Stars: 533
Hardware match: A 16–32 GB NVIDIA GPU on Windows or Linux. ExLlamaV3 / EXL3. Not a DGX Spark guide.
Second source: Mia’s model catalog

Open source recipe

The author did not publish a one-line install command. The steps are on their page.

Before you start

A 16–32 GB NVIDIA GPU
Their one-click installer for the platform you are on
The quant their README picks for your VRAM

Steps live at the source. We don’t maintain the install script.

The recipe belongs to MiaAI-Lab.

Stacks News