Meta Launches Llama 2 AI Models for Free Commercial Use
Meta releases Llama 2, a significantly upgraded family of open-source text-generating AI models available for both research and commercial applications. The new models offer improved performance and optimized compatibility across major cloud platforms and mobile devices.
Meta introduces Llama 2, a new family of AI models designed to power modern chatbots and text-generation applications. Unlike the previous Llama generation, which required special access and eventually leaked online, Meta makes Llama 2 completely free for both research and commercial use. The tech giant partners with major platforms like AWS, Azure, and Hugging Face to host these pretrained models for developers.
The Llama 2 lineup includes both base models and Llama 2-Chat variants fine-tuned specifically for two-way conversations. These models come in three sizes with 7 billion, 13 billion, and 70 billion parameters, allowing developers to choose the right level of sophistication for their specific needs. Meta trains these models on two trillion tokens of publicly available, mostly English web data that focuses on factual text.
Meta optimizes Llama 2 for wider hardware accessibility through expanded partnerships with Microsoft and Qualcomm. The models run natively on Windows and will eventually arrive on smartphones and PCs equipped with Qualcomm's Snapdragon chips in 2024. This broad availability and hardware optimization position Llama 2 as a highly competitive, open alternative to proprietary systems like ChatGPT and Google's PaLM 2.