Meta Plans Year-End Launch of Llama 4.X AI Model

Meta Platforms is accelerating the development of its next-generation AI model, internally known as Llama 4.X or Llama 4.5, targeting a release by the end of 2025. This effort is led by the Meta Superintelligence Labs (MSL), a newly established unit focused on advancing AI capabilities, with a specialized team called TBD working to improve upon the underwhelming performance of the earlier Llama 4 models, which struggled with coding, reasoning, and instruction-following tasks. CEO Mark Zuckerberg confirmed progress on intermediate versions Llama 4.1 and 4.2 and emphasized ongoing work on future models aimed at pushing AI frontiers. Despite challenges including delayed launches like the Behemoth model and some talent departures, Meta has invested heavily in AI, acquiring a significant stake in Scale AI and appointing Alexandr Wang to lead MSL. Wall Street remains optimistic about Meta's AI prospects, as reflected in a strong buy consensus and a price target implying substantial upside. The company continues to reorganize its AI operations around training, research, product, and infrastructure to maintain competitiveness with rivals such as OpenAI, Google, and Anthropic.