Meta, at its AI Day event held in London on Tuesday, confirmed that its next open-source large language model, Llama 3 (Large Language Model Meta AI), will launch within the next month. This model will power various generative AI assistants, and the same model will be used to power multiple products across Meta’s portfolio, suggests a report by TechCrunch.
At the event, Nick Clegg, Meta’s president of global affairs, stated, “Within the next month, actually less, hopefully in a very short period of time, we hope to start rolling out our new suite of next-generation foundation models, Llama 3, which will have a number of different models with different capabilities, different versatilities.”
Meta announced Llama 2 back in July 2023, which is also the first version made available to the public, and it is an open-source model. With Llama 3, the company is said to have fixed some of the drawbacks of the previous model along with the introduction of new functionalities.
Compared to Llama 2, Llama 3 is said to be a much bigger model, possibly with over 140 billion parameters, which makes it a more capable model, possibly with better reasoning. It is also said to be capable of answering questions more accurately and can also handle a wide range of questions, including controversial ones.
Like the previous generation models from Meta, Llama 3 is also said to be an open-source model, where developers can access the model for free and can incorporate it within their solutions. However, Meta’s text-to-image/video generation tool, Emu, is still limited to select users.