Brave Leo AI

Ask Leo to summarize a webpage, doc, or PDF. Or ask almost any question in general, and get answers, summaries, clarifications, related information, and other viewpoints. Brave Leo can also create content like articles, essays, and emails. It can translate between different languages, and even offer coding help by offering suggested and sample code for software applications and database queries. On iOS, you can also ask Leo questions without typing on your phone via voice-to-text.

Brave Leo is currently available in the Brave browser for desktop (macOS, Windows, and Linux) and mobile (Android and iOS).

We’ve added Mixtral 8x7B as the default LLM for both the free and premium versions of Brave Leo. We also offer Claude Instant from Anthropic in the free version (with rate limits) and for Premium. The free and premium versions of Leo also feature the Llama 2 13B model from Meta.

With respect to the servers that power Brave Leo, your browser will share your latest prompt, the context of your current conversation, and—when the use case requires it—the necessary context from the page you’re viewing. If you’re using Leo with a Brave-hosted model, once a chat is closed, all records of that chat are erased. We do not collect identifiers that can be linked to you (such as IP Address), and no personal data is retained by Brave-hosted AI models. Note that Brave Leo offers several models to choose from. If you select a model from Anthropic and submit Leo queries, that data will be processed by Anthropic for a period of 30 days, then deleted.

Brave Leo may be more or less accessible depending on traffic levels and usage. You can subscribe to Brave Leo Premium to get higher usage limits.

No account or signup is required to access the free Brave Leo experience. Leo Premium subscribers will have to create a Premium account, but this is only to verify premium access. There is no Leo account per se, and the premium experience remains private. Your premium account is completely unlinkable to any usage of Leo itself.

No. Your conversations are not persisted and they are discarded immediately after a response has been generated.

Brave-hosted models refer to Large Language Models (LLMs) hosted exclusively on Brave’s servers. If you’re using Leo with Brave-hosted models, there is no way to review past conversations or delete that data—it isn’t stored in the first place. Brave-hosted model prompts and responses are immediately discarded remotely, and discarded locally when a chat is closed. Note that Brave-hosted models are the default experience in Leo. However, Leo also offers users the ability to choose other, non-Brave hosted models, including ones from Anthropic. For users who choose an Anthropic model, Brave has no access to these conversations after a response is made to the client, but Anthropic will retain these conversations for a period of 30 days. In a future release, we hope to offer an option to return to previous conversations with on-device persistence.

As with any AI, Leo’s responses may contain misleading information and/or factual inaccuracies. Additionally, bias and/or harmful content present in the data during training may on some occasions be reflected in the model’s outputs. We recommend double-checking any Leo responses for accuracy. Please also provide feedback to us if you see anything concerning. To do so, click the menu next to Leo’s response and select the best option.

Yes, it does. Hallucinations are an intrinsic challenge in how LLMs work. Sometimes regenerating the answer may help. Always double-check Leo’s responses (for example, try the same query on Brave Search) before quoting them.