Meta Platforms (NASDAQ:META) is likely to release the next large language model in the Llama model family, Llama 4, later this month, The Information reported.

The release of Llama 4 has been pushed back at least twice due to performance issues, with underperformance in reasoning and math tasks, the news outlet added, citing two people familiar with the matter. There are also concerns that Llama 4 is less capable than OpenAI's models in having humanlike voice conversations, the people added.

Meta unveiled its most recent Llama model, Llama 3.3, in December. 

Several of Meta's competitors have released new versions of their large language models in recent weeks, including Google (GOOGL) (GOOGL), which released Gemini 2.5 on March 25. 

Meta is also discussing offering a Llama API for businesses as part of an internal program known, Llama X. This could let companies use Llama more easily in their own software and provide Meta with added revenue, the news outlet added, citing three people familiar with the matter.

Morgan Stanley said in March that the release of Llama 4 could come in the Spring of 2025. The company is also gearing up for its LlamaCon developer event next week, where it is expected to provide updates on the Llama family of models, along with other AI innovations,

Meta did not immediately respond to a request for comment from Seeking Alpha.

Chief Executive Mark Zuckerberg said in October 2024 that the company's upcoming models are “being trained on a cluster that is bigger than 100,000 [Nvidia] H100s or bigger than anything that I've seen reported for what others are doing.”

Last month, Meta announced that its Llama artificial intelligence models have been downloaded more than 1B times.