AMD and Unsloth Bring NPU Support to 100+ Agentic AI Applications AMD and Unsloth are enabling more than 100 agentic AI applications and workflows to run local inference on AMD XDNA 2 NPUs in supported Ryzen AI PCs, the first NPUs supported by Unsloth Desktop. The Unsloth Desktop API exposes an OpenAI-compliant endpoint, with Lemonade and FastFlowLM handling model serving and inference on the NPU, and supported tools include Hermes Agent, OpenClaw, Open WebUI, Cline, LangChain and GitHub Copilot. AMD said the setup gives users a power-efficient alternative to GPU execution while leaving CPU and GPU resources available for orchestration and other applications. What’s the News? AMD and Unsloth are enabling more than 100 agentic AI applications and workflows to access local NPU inference on supported AMD Ryzen™ AI PCs. AMD XDNA™ 2 NPUs are the first neural processing units NPUs supported by Unsloth Desktop https://unsloth.ai/docs/desktop , with its API that connects compatible applications to supported models running on the NPU. Why It Matters? AI assistants and agents that work throughout the day need efficient inference alongside everyday PC tasks. This collaboration gives users an alternative to GPU execution, putting AMD XDNA 2 NPUs to work on sustained, low-power inference while leaving CPU and GPU resources available for other applications. The integration builds on AMD’s ongoing work to expand and optimize software support, bringing dedicated NPU acceleration to more of the tools people use. What Is AMD’s Role? With support for Unsloth Desktop, AMD is expanding the supported compute resources for developers and users to deploy and run agentic workflows locally, efficiently and through familiar applications. What Does Unsloth Enable? Unsloth spans model development and deployment, with open-source tools that accelerate fine-tuning and reduce GPU memory requirements, alongside a desktop app for training and inference without writing code. AMD NPU support extends that ecosystem beyond CPU and GPU execution, giving users a power-efficient inference option and connecting it to more than 100 compatible agentic applications and workflows. How Does It Work? A compatible application sends a model request to the Unsloth Desktop API, while Lemonade and FastFlowLM handle model serving and inference on the AMD XDNA 2 NPU. The model’s response returns to the application for use in its workflow. The NPU accelerates supported model inference, while the application’s other functions continue to use the PC’s CPU and GPU as needed. What NPU-Powered Tools Are Available Through Unsloth? The Unsloth Desktop API offers an OpenAI-compliant endpoint for compatible apps. That integration includes more than 100 agentic AI applications and workflows. Tools include Hermes Agent, OpenClaw, Open WebUI, Cline, LangChain and GitHub Copilot. How Does It Benefit PC Users? AI agents can carry out multistep tasks across connected tools. Access to local NPU inference gives compatible agents a more power-efficient way to process model requests as they work through those tasks, leaving CPU and GPU resources available for orchestration and other applications, helping users spend less time managing individual steps and more time reviewing results. More: Learn about how AMD and Unsloth work together and how to install Unsloth Desktop on a PC that includes an AMD XDNA 2 NPU at “ AMD x Unsloth: Unlocking the World's First NPU for Unsloth Desktop https://www.amd.com/en/blogs/2026/unlocking-the-amd-ryzen-ai-npu-for-unsloth-desktop.html .” Press inquiries: corporate.pressinquiry@amd.com mailto:corporate.pressinquiry@amd.com