Meta released Llama 2, its large language model with an open source license for research and commercial use.
Microsoft is the preferred partner for Llama 2, however Meta announced support across a broad set of companies, including AWS, HuggingFace, Intel, AMD, Nvidia, IBM Qualcomm, data cricks, Deutsche Telekom, Orange, YCombinator, Palo Alto Networks, Andreessen Horowitz, MIT, etc.
The Llama 2 pretrained models have undergone extensive training with a staggering 2 trillion tokens and possess twice the context length compared to its predecessor. Furthermore, Meta says its models have been refined through rigorous training involving over 1 million human annotations. The Llama 2 models are available in three parameter sizes comprising of 7B, 13B, and 70B.