Unsloth Now Supports Windows ARM64 for Local LLMs

September 20, 2026

Unsloth Releases Windows ARM64 Binary Version

At a Glance

Item Value
Repository unslothai/unsloth
Version Windows-ARM64
Published 2026-09-16
License Apache-2.0
Source type Primary source (the publisher itself)

Values determined by this site’s code at collection time. Dates are JST.

Overview

unslothai/unsloth is a project that provides a UI for running and fine-tuning LLMs (Large Language Models) and diffusion models in local environments. It is offered in three ways: Unsloth Desktop (a desktop app), Unsloth Studio (a Web UI), and Unsloth Core (a code-based approach).

As technical features, in addition to supporting formats like GGUF and MLX, it widely supports a diverse range of model architectures such as Qwen3.8, DeepSeek-V4, MiniMax-H3, Gemma 4, and FLUX. This allows developers to operate advanced generative AI models in their own local environments.

In this release, binaries for Windows ARM64 environments are newly provided.

Key Changes

Provision of Windows ARM64 Binaries

The primary change in this update is the release of binaries for Windows ARM64. This enables users utilizing Windows devices with ARM architectures to leverage Unsloth’s features.

This change affects engineers using Windows machines equipped with ARM-based processors. It expands the options for running and fine-tuning models in ARM64 environments within workflows that previously assumed x86_64 environments.

Supported Models and Hardware

With the provision of this binary, hardware adopting the Windows ARM64 architecture is now officially supported. Consequently, engineers using Windows devices equipped with ARM-based processors can run and fine-tune the following model architectures through Unsloth:

  • GGUF
  • MLX
  • Qwen3.8
  • DeepSeek-V4
  • MiniMax-H3
  • Gemma 4
  • FLUX
    and more

How to Update

The installation procedure for developers (Developer/Nightly/Experimental) building from the latest source (main branch) in a Windows PowerShell environment is as follows:

git clone https://github.com/unslothai/unsloth.git
cd unsloth
Set-ExecutionPolicy -Scope Process -ExecutionPolicy RemoteSigned
.\install.ps1 --local
unsloth studio -p 8888

To update an already installed environment to the latest state, run the following commands:

cd unsloth; git pull
.\install.ps1 --local
unsloth studio -p 8888

Additionally, advanced installation options are available depending on the use case:

Specifying the Installation Location By using the UNSLOTH_STUDIO_HOME environment variable, you can install to an isolated location.

$env:UNSLOTH_STUDIO_HOME='C:\path';.\install.ps1 --local

Skipping PyTorch (GGUF-Only Mode) To install in a mode that uses only the GGUF format without going through PyTorch, specify UNSLOTH_NO_TORCH=1.

$env:UNSLOTH_NO_TORCH=1; irm https://unsloth.ai/install.ps1 | iex

Specifying the Backend You can force-specify backends such as Vulkan, CPU, CUDA, or ROCm during installation by using UNSLOTH_LLAMA_CPP_BACKEND.

$env:UNSLOTH_LLAMA_CPP_BACKEND="vulkan"
irm https://unsloth.ai/install.ps1 | iex

Pinning the Python Version To install by specifying a particular Python version, use UNSLOTH_PYTHON.

$env:UNSLOTH_PYTHON='3.12'; irm https://unsloth.ai/install.ps1 | iex

Specifying the npm Registry If you use a proxy in an enterprise environment or similar, you can set UNSLOTH_NPM_REGISTRY.

$env:UNSLOTH_NPM_REGISTRY='https://artifactory.example.com/api/npm/npm/';.\install.ps1 --local

Uninstallation To uninstall on Windows (PowerShell), run the following command:

irm https://raw.githubusercontent.com/unslothai/unsloth/main/scripts/uninstall.ps1 | iex

Related Articles

Sources

Update History

  • 2026-09-20: Rewrote the article from re-collected sources and restored it from draft to published.

Engines and ToolsARM64,GGUF,LLM,Unsloth,Verified,Windows