Meta announced it’s open-sourcing its large language model LLaMA 2, making it free for commercial and research use and going head-to-head with OpenAI’s free-to-use GPT-4, which powers tools like ChatGPT and Microsoft Bing.
Meta announced the move as part of Microsoft’s Inspire event, noting its support for Azure and Windows and a “growing” partnership between the two companies. At the same time, Microsoft revealed more details about the AI tools built into its 360 platform and how much those will cost. Qualcomm also announced it is working with Meta to bring LLaMa to laptops, phones, and headsets starting from 2024 onwards for AI-powered apps that work without relying on cloud services.
Meta’s press release explains the decision to open up LLaMA as one giving businesses, startups, and researchers access to more AI tools and allowing for experimentation as a community. According to Meta, LLaMa 2 was trained on 40 percent more data when compared to LLaMa 1, which includes information from “publicly available online data sources.” It also says it “outperforms” other LLMs like Falcon and MPT when it comes to reasoning, coding, proficiency, and knowledge tests.
“We believe an open approach is the right one for the development of today’s AI models, especially those in the generative space where the technology is rapidly advancing,” Meta said in a statement. “Opening access to today’s AI models means a generation of developers and researchers can stress test them, identifying and solving problems fast, as a community.”
The open-sourced LLaMa 2 will be available through Microsoft’s Azure platform. Meta said LLaMA will also be available through AWS, Hugging Face, and other providers.
While Meta first announced its LLaMA model in February, it leaked on 4chan just days later. In opening up LLaMA, Meta said it wants to improve safety and transparency. The company said LLaMA 2 model has been “red-teamed” or tested for safety by “generating adversarial prompts to facilitate model fine-tuning” both internally and externally. Meta also discloses how the models are evaluated and tweaked.