Meta Delays Launch Of Llama 4 Behemoth AI Model Over ‘Capability Concerns’

Meta Platforms is postponing the launch of its flagship Behemoth AI model reportedly because of doubts about its capabilities. According to a report, Meta’s team is questioning the capabilities of its Behemoth large-language model, and whether the enhancements are substantial enough to warrant release to the public.

Behemoth was internally planned for release in April to align with Meta’s first AI conference for developers, LlamaCon, but its deployment was then postponed till June. The recent report suggests it could only see a release in fall or later.

Notably, Meta had described Llama 4 Behemoth as “one of the smartest LLMs in the world and our most powerful yet to serve as a teacher for our new models” when it announced its preview in April. At that time, the company unveiled a smaller model, Llama 4 Scout, and a bigger one, Llama 4 Maverick, the most recent iterations of the LLM Llama.

According to the company, Behemoth has 288 billion active parameters, 16 experts and a total of 2 trillion parameters.

LLMs are the power behind AI chatbots such as Meta AI, ChatGPT, or Gemini, and are able to comprehend a human query and produce text that is human-like.

As part of its foray into AI, Meta has been releasing AI models regularly, which drive the Meta AI assistant integrated into WhatsApp, Instagram Direct, Messenger, and the web. Meta AI can be used on these social media platforms to write posts and captions and edit photos. Last month, Meta also unveiled a stand-alone app for Meta AI that also serves as a hub for its Ray-Ban smart glasses.

With competition to develop sophisticated and reasonably priced AI models rising at a fierce pace, delay in the release of new Llama models like Behemoth might cause the company to lag behind other major players like Google and OpenAI. Even as it continues to refine its existing models, Meta will be rightly concerned about how far ahead its rivals may go.