Brave is Opening Leo to LLMs and On-Device SLMs

Brave today announced it will provide Bring Your Own Model (BYOM) functionality in its flagship web browser for use with its Leo AI assistant.

“Bring Your Own Model (or BYOM) is an optional new way of using Leo, Brave’s native browser AI,” the announcement post explains. “It allows users to link it directly with their own AI models. BYOM allows users to run Leo with local models running safely and privately on their own machines. Users will be able to chat or ask questions on webpages (and docs and PDFs) to Leo without their content ever having to leave the device. [And] BYOM isn’t limited to just local models. Users will also have the flexibility to connect Leo to remote models running on their own servers, or that are run by third parties (like ChatGPT).”

BYOM is the first step towards implementing Brave’s broader plans to provide AI capabilities that work with personal data while safeguarding user privacy. By letting its customers configure their own AI models, locally and/or remotely, each can customize the AI’s behavior, output, and capabilities to their specific needs, the company says.

In keeping with its privacy focus, BYOM requests bypass Brave entirely, whether they’re local or remote.

“Brave does not act as an intermediary and has absolutely no access to—or visibility into—the traffic between the user and the model,” Brave notes. “This direct connection ensures end-to-end control for the user.”

In its announcement post, Brave explains how easy it is to configure a third-party AI model for use with Leo. If it’s a local SLM, just download and install the model—Brave recommends Ollama Llama 3, Mistral AI’s Mistral 7B, and Microsoft Phi 3 Mini for their combination of small size and capabilities—and then configure it for use with Leo in Brave settings. If it’s a cloud-hosted LLM, you just need its URL, desired model name, and private API key.

For now, BYOM is limited to text input and output, but it appears that will change in the future. If you’d like to get started testing this interesting new feature, it’s available now in the Brave Nightly channel builds.