Meta, Microsoft unveil next-gen open source large language model Llama

Meta and Microsoft have announced the availability of Llama 2, the next generation of their open-source large language model. Llama 2 represents a significant milestone in the development of AI technologies, offering advanced capabilities and new opportunities for researchers, developers, and businesses worldwide. By opening access to Llama 2, Meta and Microsoft are fostering open innovation and collaboration in the field of generative AI.

Llama 2, which is now free for both research and commercial use, builds upon the success of its predecessor, Llama 1, which garnered over 100,000 requests for access. With Llama 2, the tech giants are providing the model weights and starting code for the pretrained model, as well as conversational fine-tuned versions. These resources empower users to explore and build upon the large language model's capabilities.
As part of this partnership, Microsoft becomes the preferred partner for Llama 2, underscoring our commitment to driving innovation and empowering developers to create cutting-edge AI applications. Llama 2 is available in the Azure AI model catalog, allowing developers using Microsoft Azure to leverage their cloud-native tools for content filtering and safety features. Additionally, Llama 2 is optimized to run locally on Windows, offering developers a seamless workflow across multiple platforms. It is also accessible through other providers such as Amazon Web Services (AWS) and Hugging Face.
By adopting an open approach to AI model development, Meta and Microsoft aim to democratize access to foundational AI technologies and drive progress across industries. We believe that making AI models available openly benefits everyone, enabling businesses, startups, entrepreneurs, and researchers to experiment and innovate in exciting ways. This approach also promotes safety and accountability by fostering a community of developers and researchers who stress-test the models, identify vulnerabilities, and collectively work towards responsible AI development.

Meta has a long-standing commitment to open source, exploratory research, and collaboration with academic and industry partners. Our dedication to an open ecosystem has led to the release of numerous large language models, which have fueled progress in generative AI-powered experiences. We are thrilled to witness the amazing achievements made by researchers and businesses using Llama 1 and are excited to see the groundbreaking projects that will emerge with the introduction of Llama 2.
Responsibility is a core focus for Meta and Microsoft. We recognize that AI technologies carry risks and are committed to building AI models responsibly. To that end, we are providing a range of resources to support users of Llama 2. These resources include red-teaming exercises, transparency schematics, a responsible use guide, and an acceptable use policy. By fostering transparency, sharing knowledge, and setting guidelines, we aim to ensure the safe and responsible deployment of large generative models.
In addition to technical support, Meta has launched new initiatives to harness the collective wisdom and creativity of the global community. The Open Innovation AI Research Community invites academic researchers to join a program dedicated to responsible development and sharing of large language models. This collaborative platform will facilitate knowledge sharing and contribute to ongoing research in this vital area. Furthermore, the Llama Impact Challenge aims to inspire a diverse range of entities to utilize Llama 2 in addressing critical challenges in fields such as the environment and education.
As pioneers of open-source frameworks, such as React and PyTorch, Meta has witnessed firsthand the transformative power of openness in technology development. By openly sharing today's large language models, we aim to support the advancement of helpful and safer generative AI. We are excited to witness the innovative solutions and applications that the world will create with Llama 2.