Brave boosts Leo AI Assistant with Next-Gen Llama 3

Brave browser has quietly upgraded its language models to reflect significant changes. Brave has ditched old Claude Instant and upgraded Llama 2 13b to Llama 3 8b. Mixtral remained the default AI assistant for Leo users. The changes are live in the Nighly version of Brave and will be reflected across apps and stable versions in other platforms in the future.

Brave’s built-in AI Assistant: Leo

Leo, Brave’s built-in AI assistant, can answer your questions, generate page summaries and GitHub pull requests, and translate content across languages.

Leo’s present Language Models:

Mixtral: This is the default model for Leo in the Brave browser and excels at advanced chat tasks. It can handle complex queries, provide detailed summaries, and deliver nuanced responses. In other words, for in-depth information or complicated questions, Mixtral is your go-to LLM that can dig deep and find the best answers.

Claude Instant: This is used for conversation and text generation. When you want to have a casual conversation or have Leo write something short for you, Claude Instant is your chat assistant.

Claude 3 Haiku: This lightning-fast language model prioritizes rapid responses in chat environments.
(Note: Claude Instant has been removed)

Claude 3 Sonnet (Premium): This model (likely paid) can help you with answering questions.

Llama 2 13B: serves as a general-purpose chat companion.

Brave upgrades Leo AI’s language models

Brave has upgraded models and left Claude Instant as we spotted in the Nightly version.
Mixtral continues to remain the default and is available to excel at complex chat tasks like before.

While Claude Instant is gone, Claude 3 Haiku (free) and Claude 3 Sonnet (premium) LLMs are still available to assist you with questions.

Brave has replaced Llama 2 13b with Meta’s next-generation and powerful language model, Llama 3 8b. Llama 3 is expected to be more powerful than its predecessor, potentially offering superior performance in various tasks like question answering and creative text generation.

Removing an LLM like Claude Instant simplifies the user experience. Users won’t need to choose between different models for basic tasks.

All in all, the upgraded Leo promises improved performance, simplified interaction, and streamlined experience where Llama 3 8B is likely more efficient than the combined power of Llama 2 13B and potentially Claude Instant. This could translate to faster response times and improved accuracy.