Meta Platforms has reportedly postponed the release of its highly anticipated AI model, Behemoth, due to internal concerns over its performance, according to a Wall Street Journal report citing sources close to the matter.
Engineers at the tech giant are said to be struggling to achieve major advancements in Behemoth’s capabilities, prompting internal debate about whether the latest version shows enough progress over its predecessors to warrant a public launch.
Initially, Meta had planned to unveil Behemoth in April during its first-ever AI developer conference. The launch was later shifted to June but has now been delayed again—this time potentially to the fall or even later, according to the report.
Meta has not issued an official statement in response to requests for comment on the delay.
The Llama 4 Scout and Llama 4 Maverick are two new versions of the Llama 4 models that Meta unveiled back in April. At the time, the company described Llama 4 Behemoth as its most advanced large language model (LLM) to date, highlighting its potential to serve as a “teacher” for future models.
Despite the hype, the delay signals the challenges even tech giants face in refining cutting-edge AI systems to meet increasingly high standards for public deployment.