Meta has once again delayed the release of its advanced artificial intelligence model, Behemoth, originally expected to debut in April. According to a report by The Wall Street Journal on May 15, the company initially postponed the launch to June but has now pushed it back until at least the fall. Sources familiar with the matter indicated that the delay stems from ongoing efforts to improve the model and doubts about whether its performance will match the company’s earlier claims.
Despite these internal concerns, Meta has publicly stated that Behemoth surpasses models developed by major competitors like Anthropic, Google, and OpenAI in certain benchmarks. On April 5, when Meta introduced the latest versions of its Llama AI models, it also previewed Behemoth, describing it as one of the most intelligent and powerful language models the company has created to date.
During Meta’s first-quarter earnings call on April 30, CEO Mark Zuckerberg announced plans to significantly increase the company’s investments in AI infrastructure. He said Meta would boost capital spending to between $64 billion and $72 billion in 2024, up from the earlier estimate of $60 billion to $65 billion. This funding will support the growing demand for AI computing capacity across Meta’s platforms, including Facebook, Instagram, and WhatsApp.
Zuckerberg emphasized the growing role of AI in shaping Meta’s future, stating that the company is accelerating efforts to expand its data center capabilities. These upgrades aim to accommodate both immediate and long-term AI development needs.
Meanwhile, Meta’s April 5 release of open-source models Llama 4 Scout and Llama 4 Maverick positions the company as a cost-effective alternative to competitors with proprietary systems. These models are seen as beneficial for businesses looking to deploy AI without the expense associated with closed platforms.
