Meta to release its new AI model to the public

Meta has released a new AI model, Llama 3, which it claims to be more powerful than OpenAI’s ChatGPT and Google’s Gemini.

The tech giant will open-source the new AI to allow developers to build products using the model for free. However, big cloud giants including Apple and Google will be charged for training their models on the system.

The parent company of Facebook, Instagram, and Whatsapp said that Llama 3 has “vastly improved capabilities”, such as the ability to reason, and it aims to make it multilingual and multimodal in the future.

Meta explains that the model’s “state-of-the-art performance” derives from its 8B and 70B parameter models, making it seven times larger than its predecessor, Llama 2 – This essentially means there is a greater number of variables used to train the systems and shape its output, allowing it to perform better than smaller models.

“We are delighted to pen source so that developers, entrepeneurs and researchers can feed off Llama 3 for new innovation,” head of global affairs at Meta, Nick Clegg, told the Financial Times. “[but not for] very large platforms who might just want to piggyback on our investments to sort of short circuit their own.”

“We’re not pouring billions of dollars into this stuff in order to save Apple money,” Clegg added.

Llama 3 is available under its chatbot ‘Meta AI’ for public use on Instagram and Facebook search bars, on the web, and as a chat on Whatsapp for the public to use.

However, Clegg explained that it is not available in the UK and Europe yet, “but [are proceeding] at the pace of the discussions we have with regulators and others.”