AI & Agents

Unsloth Desktop Adds Local Inference on AMD Ryzen AI NPUs, Including on Linux

Leave a comment

Unsloth Desktop now supports local inference on AMD Ryzen AI NPUs, Phoronix reports, starting with XDNA 2 and later. AMD confirmed to Phoronix that the support works with the Linux builds, not only Windows. The NPU path uses AMD’s open-source Lemonade and FastFlowLM code for model serving and inference. AMD and Unsloth say more than 100 compatible agentic applications and workflows can already reach those NPU-hosted models through Unsloth’s API. AMD’s blog and Unsloth’s own docs list supported hardware in the Ryzen AI 300, Ryzen AI Max, and Ryzen AI 400 families with an XDNA 2 NPU, on Windows 11 or Linux x64. Linux needs a recent kernel with the amdxdna driver, or an older kernel with amdxdna-dkms, plus current NPU firmware.

This is NPU inference, not a claim that every model will suddenly run fast on every Ryzen laptop. Memory requirements still matter, and I have not tried Unsloth Desktop on an NPU for this post.

Review before you install: read the release notes yourself, back up your config, and try new versions somewhere you can afford to break first.

My take

Local models on the NPU, with Linux included, is the part that matters to me. If you already have an XDNA 2 Ryzen AI box, this is the official path to try, after you read the hardware and driver notes. I will not pretend I have already run it.

God Bless, and May the Source Be with You!

References & Further Reading

Subscribe (RSS) Follow AI & Agents via RSS

Leave a note

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.