Meta He highlighted the adoption of his Llama language model family, which is approaching 350 million downloads, a success that he identifies with his commitment to open source.
Just over a month ago Meta released Llama 3.1, a version of its foundational model that matches the most advanced closed rivals with its 405 billion parameters (405B) and skills in general knowledge, orientation, mathematics, use of tools and multilingual translationIt also supports a context window of 128,000 tokens.
With Llama 3.1, Meta offers developers “a complete reference system to more easily create their own custom agents along with a new set of security and protection tools to help them develop responsibly,” as the company noted in its assessment of the adoption of its models.
Together, the Llama models – the first of which appeared in February last year – have registered nearly 350 million downloads, Of which 20 million are limited to the last month which, according to Meta, makes Llama “the leading open source family of models.”
“By making our Llama models available to the public, we have seen a dynamic and diverse AI ecosystem come to life where developers have more choice and capabilities than ever before.“, the company said in its assessment of the adoption of its models.
The company also shares that Llama’s monthly usage increased tenfold between January and July 2024. for some of the cloud service providers. And with the release of Llama 3.1 they have increased the number of partners in the Llama early access program by five times.