Jan provides a streamlined experience for running both local and cloud-based artificial intelligence models on your desktop, supporting platforms like Mac, Windows, and Linux. To get started, users first download and install the Jan application for their preferred operating system. Upon launching Jan, the user interface appears without any pre-installed models, allowing for a clean slate tailored to individual needs. From there, users may either download local models directly or connect to external cloud-based artificial intelligence providers.
To add a local artificial intelligence model, users navigate to the Hub Tab within Jan or, alternatively, browse the Hugging Face model repository. The platform offers a broad spectrum of options, from lightweight models suitable for modest hardware to resource-intensive powerhouses. Notably, all models must be in the GGUF format for compatibility with Jan. Users can review model details, ensure their hardware meets minimum requirements, and initiate the download. For some Hugging Face models, an access token may be needed—this can be configured via Jan´s settings under model providers.
Jan accommodates performance optimization through optional GPU acceleration, which can be enabled for compatible Windows and Linux systems. This requires installing the necessary drivers and adjusting hardware settings within the app. Once a model is ready, users can customize assistant instructions to guide how responses are generated, managing these configurations in the Assistant section. Jan also enables seamless switching between models and fine-tuning of model parameters for different conversations. For advanced workflows, integration with cloud models from providers like OpenAI, Anthropic, Groq, and Mistral is supported—requiring only a valid API key to activate these services. With user-controlled setup, customization, and broad compatibility, Jan empowers users to harness artificial intelligence models locally and through cloud platforms in a user-friendly environment.
