Home / News / Technology / Meta’s Llama 3 Release Date – What to Expect From Open-Source AI Model?
3 min read

Meta’s Llama 3 Release Date – What to Expect From Open-Source AI Model?

Last Updated April 11, 2024 8:22 AM
James Morales
Last Updated April 11, 2024 8:22 AM

Key Takeaways

  • Meta has announced that it will release a small version of LLaMA-3 within the next month.
  • The firm intends to release a larger version of the foundation model later in the year.
  • Unlike other AI developers, Meta lets people use its models for free.

On Tuesday, April 9, Meta confirmed that it plans to release a light version of LLaMA-3 within the next month.

The latest addition to the LLaMA family is expected to be open-source, or at least, as open-source as its predecessors. Meanwhile, larger, multi-modal versions of the model are anticipated later in the year.

LLaMA 3 in 3 Sizes

Following a precedent set by Google’s Gemini and Anthropic’s Claude 3, the latest generation of LLaMA models will reportedly  be available in 3 different sizes, with the first 2 expected to be released imminently. 

Appearing the confirm the rumors, Meta’s vice president of global affairs, Nick Clegg, told a recent event in London that “in a very short period of time, we hope to start rolling out our new suite of next-generation foundation models.”

Throughout the year, “there will be a number of different models with different capabilities [and] different versatilities,” he added.

New Model to be Integrated Across Meta Products

According to Meta Chief Product Officer Chris Cox, LLaMA-3 will be used to power a range of new AI services across the company’s products.

“Our goal over time is to make a LLaMA-powered Meta AI be the most useful assistant in the world,” Vice President for AI Research Joelle Pineau added.

While Meta has been tight-lipped on specifics, reading between the lines Pineau’s comment suggests the company may be preparing to launch its own chatbot, which could potentially be offered via one of its social media platforms.

Of course, there are already plenty of LLaMA-powered chatbots available. But so far, Meta has abstained from building one itself

Meta’s “Open-Source” Approach to AI

Unlike other leading AI developers, Meta has not closed off its foundation models behind paywalls.

Whereas equivalent proprietary models developed by OpenAI and Google can only be accessed via Application Programming Interfaces (APIs),  Meta has made trained LLaMA models and weights freely available under a permissive software licenses.

llama, open-source AI
Although ostensibly open-source, LLaMA-2 falls short of the standard of openness many researchers want. Source: opening-up-chatgpt.github.io

That being said, critics have pointed out that the AI isn’t fully open-source as the all-important training and instruction-tuning data remains a closely guarded secret.

As noted  by the Institute of Electrical and Electronics Engineers, “while third parties have been able to create applications that extend on the base model, aspiring developers and researchers have a limited ability to pick apart the model as is.”

Was this Article helpful? Yes No