Meta is once again delaying the release of its large-scale generative AI model called “Behemoth,” which was initially set to debut in June. According to The Wall Street Journal , the engineering team has encountered difficulties in enhancing the model’s capabilities — the release may be postponed to fall or even later. This is causing quite a stir, as the company recently claimed that “Behemoth” already surpasses all major competing solutions, including GPT-4.5, Claude Sonnet 3.7, and Gemini 2.0 Pro, in a number of technical tests in the STEM field.
Internal sources indicate that despite ambitious claims, “Behemoth” does not show a significant leap in quality compared to existing models. Meta even has doubts about whether it should be publicly presented anytime soon. Meanwhile, the company has already released the fourth generation of its Llama lineup: currently, users have access to two smaller versions — “Scout” and “Maverick,” and another lightweight modification has been announced.
Meta is actively integrating AI features into its products — Facebook, Instagram, WhatsApp, and Messenger. Users can create posts, change chat backgrounds, and even edit images with new tools. At the end of April, the company introduced a separate application “Meta AI,” but the delay of “Behemoth” underscores that the race in the field of generative AI is becoming increasingly intense.
