Meta Releases Llama 3.1, Its Most Powerful Open Source AI Models Yet
Meta unveils Llama 3.1, featuring the 405B model that rivals top closed-source AI systems with a massive 128K context window. The release includes new safety tools and immediate support from over 25 major tech partners.
Meta releases Llama 3.1, a groundbreaking suite of AI models that includes the 405B version, the first frontier-level open source large language model. This massive model rivals the best closed-source systems on the market, offering state-of-the-art performance in general knowledge, math, tool use, and multilingual translation across eight languages.
The new models feature an expanded context length of 128K tokens, allowing developers to process much larger documents and conversations. Meta provides these models alongside new safety tools like Llama Guard 3 and Prompt Guard, while also introducing the Llama Stack API to standardize how third-party projects leverage the AI system.
Over 25 major partners, including AWS, NVIDIA, Databricks, Google Cloud, and Azure, offer immediate support for the Llama 3.1 ecosystem. Developers in the United States can test the impressive capabilities of the 405B model right now by asking complex math or coding questions through WhatsApp or Meta AI.