Llama-3.3-70B-Instruct

Llama 3.3 70B Instruct is an instruction-tuned large language model published by Meta and released on December 6, 2024 as part of the Llama 3.3 family. Meta distributes the model artifacts, including trained weights and accompanying documentation, through its official channels, with the documentation hosted at llama.com/docs/overview. The release is governed by the Llama 3.3 Community License Agreement, an open-weight license under which the foundational model code, trained weights, and supporting tooling are made available for use, reproduction, distribution, and modification, subject to the license terms.

As an open-weight instruction-tuned model in the Llama line, Llama 3.3 70B Instruct is positioned for developers and organizations that want to run a capable chat-oriented model on their own infrastructure rather than depend solely on closed APIs. Its place in the Llama 3.3 family signals an evolution focused on practical assistant behavior at the 70-billion-parameter scale, with Meta continuing to publish documentation and licensing terms that allow fine-tuning, redistribution, and integration into downstream applications. This makes it a fit for teams that value self-hosting, customization, and transparent licensing terms alongside strong general-purpose instruction following.