Meta Platforms Inc (NASDAQ:META, XETRA:FB2A, SIX:FB) is developing a new artificial intelligence model, codenamed Avocado, as a potential successor to its Llama series, according to a CNBC report.
The model, which is expected to launch in the first quarter of 2026, represents a shift toward a more proprietary approach compared with the open-source Llama family.
Sources told CNBC that Avocado had initially been expected before the end of 2025 but is now facing delays due to performance testing and training challenges.
Meta spokespersons said training efforts remain on schedule. “Our model training efforts are going according to plan and have had no meaningful timing changes,” the spokesperson told CNBC.
The move reflects broader changes at Meta, including leadership shifts and a multibillion-dollar investment in top AI talent. In June, Meta spent $14.3 billion to hire Scale AI founder Alexandr Wang and several leading engineers and researchers, with the aim of strengthening its AI capabilities and competing with rivals such as OpenAI, Google, and Anthropic.
Shares of Meta traded down 1.5% at $657 on Tuesday afternoon, up 12% so far this year.