Llama 3.1 405B model is now available for Perplexity users

Perplexity has announced that Meta’s Llama 3.1 405B large language model (LLM) is now available to access its users.

Pro subscribers can use this AI model to generate answers or set it in default. To do so, go into profile, then settings and AI Model to Llama 3.1 405B alongside recently released GPT-4o and Claude Sonnet 3.5.

The new Llama 3.1 brings 405 billion parameters as the large open-source LLM to date to provide significant upgrades in performance.

It features a new context length of 128,000 tokes equivalent to approximately 400 pages of text for advanced use cases such as long-form text summarization and complex programs.

Meta Llama 3.1 405B available on Perplexity (Source – Perplexity)

This is a multilingual model including Portuguese, Spanish, German, French, Hindi, and Thai. It is trained on 15 trillion tokens using 16,000 Nvidia H100 AI chipsets.

The new Llma model quantized from 16-bit to 8-bit numerics to run within a single server node. This technical specification enables an easy deployment while providing powerful performance.  Developers can download and customize the Llama 3.1 405B.

In addition to the web app, Perplexity has announced Llama 3.1 405B is now available to the iOS mobile app and Android will come soon.