AMD Launches Lemonade for Local AI Models on GPUs and NPUs, but Omits NVIDIA Support
Updated
Updated · InfoWorld · May 13
AMD Launches Lemonade for Local AI Models on GPUs and NPUs, but Omits NVIDIA Support
3 articles · Updated · InfoWorld · May 13
AMD has introduced Lemonade, a local AI server app with a GUI, CLI and embeddable server mode aimed at running models on AMD GPUs, Ryzen NPUs, Vulkan and CPUs.
The software leans on broad compatibility rather than deep tuning: it supports OpenAI-, Ollama-, Anthropic- and llama.cpp-style APIs, plus back ends including llamacpp, whispercpp, sd-cpp and ryzenai-llm.
Model support spans GGUF and ONNX formats, with a built-in catalog for LLMs and image models such as Gemma, Qwen, Flux and Stable Diffusion.
Its biggest limitation is hardware and interface flexibility: NVIDIA GPUs are not supported, Stable Diffusion lacks Vulkan acceleration, and the GUI exposes only basic controls like temperature, top-K and top-P.
That leaves Lemonade best suited to AMD-centric systems where models fit in memory, while users needing richer GUI controls, chat history or standard NVIDIA support may prefer alternatives like LM Studio.
Is the local AI boom a real revolution or a market bubble fueled by tools like AMD's Lemonade?
Can AMD's new AI software finally break NVIDIA's stranglehold on the local AI market?
As AI models advance, can simplified apps truly unlock the power of next-generation consumer hardware?