Llama Family is a home for llama models, technology, and enthusiasts, fostering an open platform for developers and tech enthusiasts to collaborate on the llama open-source ecosystem. From large to small models, covering various modalities and algorithm optimizations, the aim is to democratize AI for all.

By joining the Llama Family, one can progress alongside technology, community, and move towards AGI together. The meta open-source Llama model is widely used in the industry and academia, with the latest training data volume reaching 2.0t tokens and parameter sizes varying from 7b to 70b.

Additionally, the code Llama model utilizes public code datasets for training, offering base, python, and instruct model categories with parameter sizes ranging from 7b to 70b for code generation, optimization for Python, and instruction programming.

Furthermore, the Atom mega-model, a collaboration between Atom Echo and the Llama Chinese community, enhances the Chinese language capabilities of the Llama model through training on 2.7t Chinese and multilingual text data, with parameter sizes ranging from 1b to 13b.