Meta set to launch Llama 4 AI model this month

• Published On: 04 April 2025 at 20:31

Meta is set to deploy the next version of its huge language model, Llama 4, later this month, according to a report published on Friday by The Information.

However, the release could be postponed again, according to sources familiar with the situation.

Llama 4's launch has already been postponed at least twice, as Meta strives to stay up with the escalating global fight for artificial intelligence dominance. The digital behemoth, which owns Facebook, has struggled with Llama 4's performance, particularly in technical standards for reasoning, mathematics, and humanlike voice conversation—areas where competitors such as OpenAI have excelled.

Meta's substantial push into AI comes as industry heavyweights expand their investment in machine learning technologies. Under pressure from investors to demonstrate actual results, the business aims to invest up to $65 billion in building its AI infrastructure this year.

The new model is expected to incorporate certain improvements from DeepSeek, a burgeoning Chinese AI company that has developed a competitive, cost-effective alternative. Notably, one version of Llama 4 is stated to use a "mixture of experts" approach—a machine learning method in which different elements of the model are trained to specialise in specific tasks, hence increasing efficiency and accuracy.

Meta is allegedly exploring a phased release strategy for Llama 4, first incorporating it into Meta AI products and then releasing it as open-source software.

Last year, the company released Llama 3, a largely free AI model that supported eight languages, generated better code, and handled more challenging math problems than its predecessors.

Llama 4's release date has yet to be officially confirmed by Meta.