The next generation of open source large language model Llama 2 is available for free for research and commercial use.
Meta has released LLaMA 2, its first large language model that is available for anyone to use for free. The company hopes that making LLaMA 2 open source might give it an edge over rivals like OpenAI. Meta is releasing a suite of AI models, which include versions of LLaMA 2 in different sizes, as well as a version of the AI model that people can build into a chatbot, similar to OpenAI’s ChatGPT. Meta admits that LLaMA 2 still has the same problems that plague all large language models, such as a propensity to produce falsehoods and offensive language. However, the company hopes that by releasing the model into the wild and letting developers and companies tinker with it, it will learn important lessons about how to make its models safer, less biased, and more efficient.
Congratulations to Meta on a monumental day for AI and LLMs.
This is presently the most capable LLM immediately accessible as weights to anybody from academics to businesses.
The Model Revolution: Meta’s Llama v2 Makes Waves in AI World
The models seem to be extremely robust, as seen in Table 4 of the paper: The 70B model of MMLU is somewhat lower than GPT-3.5. However, HumanEval (a horrible misnomer) suggests that coding competence is much lower (48.1 vs 29.9)