Meta is reported to be delaying the release of its flagship AI model "Behemoth"

Meta Platforms (META. O) is delaying the launch of its flagship AI model, according to people familiar with the matter, sparking internal concerns about the direction of its tens of billions of dollars in AI investment, according to the Wall Street Journal. Engineers are scrambling to improve the performance of a big language model called Behemoth, with employees questioning whether it will be improved enough to support a public launch, according to people familiar with the matter. The model was originally scheduled to be unveiled at Meta's first AI developer conference in April, but it was postponed to June, and now it has been postponed again until autumn or later. Although Meta has previously been praised for catching up quickly with rivals such as OpenAI and spent $72 billion to achieve Zuckerberg's AI ambitions, the performance of the "behemoth" training has been hindered, resulting in the performance falling short of expectations. The executive team is dissatisfied with the Llama 4 development team and is considering a management reorganization of the AI product division.