Meta will delay the release of its flagship AI model, "BEHEMOTH."

On May 16th, as reported by WSJ, sources with knowledge of the matter stated that Meta Platforms (META.O) will postpone the launch of its flagship AI model. This has sparked internal concerns regarding the direction of its multi-billion-dollar AI investment. Sources indicated that engineers are working diligently to enhance the performance of a large language model named "Behemoth." Employees are questioning whether the extent of these improvements is sufficient to support a public release. The model, which was originally scheduled to make its debut at Meta's first AI Developer Conference in April, was later postponed to June and is now further delayed until the fall or beyond. Despite Meta's previous recognition for quickly closing the gap with competitors like OpenAI and its commitment of $72 billion to fulfill Zuckerberg's AI ambitions, the "Behemoth" training setback has led to underwhelming performance. The executive team is dissatisfied with the Llama 4 development group and is considering a management restructuring in the AI product division. (Jinse)