Meta confirms initial release of Llama 3 LLM within weeks

Several iterations to be rolled out over next few months

At an event yesterday, Meta confirmed its plan for an initial release of Llama 3, the next generation of its LLM, within the next month.

"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," said Nick Clegg, Meta's president of global affairs.

"There will be a number of different models with different capabilities, different versatilities [released] during the course of this year, starting really very soon."

Clegg's comment confirms the report on Monday in The Information that a limited version of Llama 3 will be available within a week or two with the full version being made available in the summer.

Speaking at the same event Meta Chief Product Officer Chris Cox said that Llama 3 will power multiple products across the Meta portfolio.

Meta has been racing to catch up with its competitors in the generative AI powered assistant field since OpenAI dropped ChatGPT on the world late in 2022. The company has spent billions of dollars on Nvidia H100 GPUs to train the model.

Llama 2 was released under open licence last summer, but the reaction probably wasn't quite what Meta hoped for. The release was criticised for having so many guardrails built in, that it infantilised users. It was an ironic outcome for a company which many people believe is guilty of perpetuating harm via other platforms like Instagram.

Last November, Meta announced the disbanding of its responsible AI team with individual team members being integrated into core generative AI teams. The redistribution of the team was undertaken to enhance responsible AI development and use.

Although it is racing to catch up with OpenAI/Microsoft and Anthropic in its LLM development, Meta gives the appearance of being reasonably cautious. The fact that Llama 2 was released on an open licence aided transparency, and a ranking by Stanford University last year put Llama 2 top of the transparency rankings.

The open licence strategy is certainly a useful one if you're trying to gain traction with the development community.

Llama 3 is expected to be approximately double the size of its previous iteration in terms of data points, and that will invariably bring the more controversial questions of potential users into scope. This might well increase its popularity with users, but could potentially create another unwelcome (from Meta's perspective) ethical debate around the biases baked into generative AI models, and associated real world outcomes.