Trending Topics

AMD claims it can run Llama 4 Scout LLM with 128 billion parameters on a Ryzen AI Max+ laptop
AMD has announced that it’s now possible to run the Llama 4 Scout local language model (LLM) with 128 billion parameters on a Ryzen-powered laptop. That’s roughly the size of ChatGPT 3.0.
“This represents a significant shift in where and how frontier AI models can be deployed,” an AMD spokesperson said, understatedly.
To look at this another way, running such a model was previously reserved for cloud-based servers, AMD claims. It hints at a momentum shift from the cloud to local AI, although much work still needs to be done.
“[Local AI] can provide greater control, reduce ongoing costs, and deliver important privacy gains by eliminating the need for third-party infrastructure,” AMD stated. “With this update, AMD enables users to run these models directly on-device, providing greater flexibility in how they deploy AI.”
It added: “This isn’t just about bringing cloud-scale compute to the desktop; it’s about expanding the range of options for how AI can be used, built, and deployed locally.”
AMD summarises this feat in three bullet points:
- The first Windows AI PC to run Llama 4 Scout 109B locally with full vision and tool use
- Support for 256,000 token context windows for long documents, multi-step reasoning and AI agent workflows
- Flexibility to run a wide range of models from lightweight assistants to advanced tools for coding and visual tasks
Why your AMD laptop probably won’t run the Llama 4 Scout LLM just yet…
Where can you find a laptop or workstation equipped with such power? AMD lists five machines in the blog announcing the feat: Asus ROG Flow Z13, Corsair AI Workstation 300, Framework Desktop, HP ZBook Ultra G1a and the HP Z2 Mini G1a.

If you happen to own such a computer, then you can run the model yourself by downloading LM Studio and the Adrenalin driver, currently in preview. More details are on the official AMD blog announcing the news.
And if you want to dig into the technical details, head to AMD’s Q&A article.
