Meta is racing the clock to ship its newest Llama AI model this year

Meta is in a race against time to launch its latest Llama AI model, Llama 4.X, by the end of this year. This ambitious project is spearheaded by a specialized team within the newly established Meta Superintelligence Labs (MSL). According to sources familiar with the development, who requested anonymity, the initiative is among the first to emerge from MSL, which is focused on advancing Meta's AI capabilities. The Llama 4.X model, sometimes referred to internally as Llama 4.5, is being prepared for production readiness to meet the anticipated release deadline. The team working on this model is also addressing existing issues with the earlier Llama 4 models, which were launched in April but received lukewarm feedback from developers. Many felt the initial models fell short in practical applications like coding and reasoning. In addition to Llama 4.X, Meta has been working on another AI model named Behemoth, which is part of the Llama 4 series. However, reports surfaced in May indicating that the rollout of this model has been delayed. A spokesperson for Meta declined to provide further comments but referenced CEO Mark Zuckerberg's remarks during the company's second-quarter earnings call, where he highlighted the progress being made on Llama 4.1 and 4.2, along with future models. Zuckerberg initially announced the formation of MSL in June, and by August, the company had reorganized its AI teams into four key areas: training, research, product, and infrastructure. Alexandr Wang, head of MSL, indicated in a subsequent memo that a subgroup named TBD would focus on training and scaling large models to enhance AI capabilities, including the development of a so-called 'omni model.' Despite the promising start, MSL has already faced some staff turnover, with at least eight employees—including researchers and engineers—departing in recent months. This comes shortly after Zuckerberg's aggressive hiring strategy aimed at attracting top AI talent from competitive organizations like OpenAI and Google DeepMind, often with lucrative compensation packages. As Meta pushes forward with its AI ambitions, the challenges of maintaining talent and delivering on expectations loom large.