Meta Unveils Llama 3.1 405B as Largest Open-Source AI Model
Meta launches Llama 3.1 405B, a massive 405-billion-parameter open-source language model backed by major cloud providers. The model features a 128k token context window, multilingual support, and advanced reasoning tools.
Meta releases Llama 3.1 405B, marking the debut of the largest open-source language model to date with 405 billion parameters. The company trains this massive model using 15 trillion tokens and 16,000 GPUs, focusing heavily on data curation, scale, and complexity management to ensure high-quality outputs. Alongside this flagship model, Meta also releases smaller 8B and 70B versions to serve a wider range of developer needs.
Major cloud vendors quickly announce their support for deploying the new Llama models across their platforms. Amazon Web Services makes the model available through Amazon Bedrock, Microsoft offers it via Azure AI’s Models-as-a-Service as a serverless API endpoint, and Cloudflare integrates the 8B version into its network. Additional launch partners include Databricks, Dell, Nvidia, IBM, Snowflake, Scale AI, and Groq, ensuring developers have widespread access to the technology.
The new Llama models boast a 128k token context window that allows users to input hundreds of pages of text at once. They also provide multilingual capabilities supporting eight languages, including English, French, Spanish, and Hindi. Furthermore, Meta equips the models with built-in tools for web search, mathematical reasoning, and code execution, making them highly versatile for complex enterprise applications.