LM Studio

LM Studio is a desktop application that makes running AI language models locally accessible without command-line knowledge or technical setup. Where Ollama requires comfort with terminal commands, LM Studio provides a graphical interface for browsing, downloading, and running local AI models, making local AI accessible to a broader audience.

The model discovery feature browses Hugging Face directly from within LM Studio, showing available models with file sizes, quantisation options, and performance characteristics. Users can filter by model family, task type, and hardware compatibility, then download directly to their computer without navigating command-line tools or understanding model file formats.

The chat interface is a familiar conversation UI similar to ChatGPT, making the local model experience immediately intuitive for users transitioning from cloud AI assistants. There is no visible distinction between the interface and cloud AI tools, which reduces the learning curve.

The local inference server, which provides an OpenAI-compatible API from a locally running model, enables developers to connect existing applications that use the OpenAI API to a local model with minimal code changes. This is useful for development and testing without incurring API costs.

Hardware requirements are the primary limitation. Smaller models like Phi-3 Mini (3.8B parameters) and Llama 3.2 3B run on most modern laptops with 8GB RAM. Models producing higher-quality outputs require more RAM and optionally a GPU. Apple Silicon Macs are particularly well-suited because the unified memory architecture allows running models efficiently without a discrete GPU.