Brave launches 'Bring Your Own Model' feature for desktop to integrate custom AI models

With the release of Brave version 1.69, desktop users can now integrate their own AI models into the Brave browser via its built-in assistant, Brave Leo. This new feature, known as "Bring Your Own Model" (BYOM), facilitates the connection of Leo with AI models running locally on users' devices or on remote servers, including third-party services like ChatGPT. This integration aims to enhance user privacy and security by allowing local models to ensure that data remains on the user's device, while remote models make direct requests from the user's device to the model endpoint, bypassing Brave entirely.

Initially available on the Brave Nightly channel for developers and testers, BYOM is scheduled for a full release in the summer of 2024, exclusively for desktop users. To set up local models, users should run an AI model on their own device using an app like LM Studio or Ollama, ensuring that no data leaves their devices. Then, to integrate these local models with Leo, users need to go to Brave’s settings, find the "Bring your own model" section, and follow the instructions to add their model.

As of August 22, 2024, BYOM is generally available to all desktop users with Brave version 1.69 or higher. The feature currently supports only textual inputs and outputs, with other modalities not yet supported. This development marks a significant step in providing users with more control over their data and the AI models they use within the Brave browser.