Despite spending billions of dollars on AI infrastructure and poaching talent from rival companies, Meta’s AI ambitions appear to have hit a snag.
The New York Times reports, citing several unnamed sources, that the social media company’s new foundational AI model is being delayed. The decision reportedly came after the model fell short in internal performance tests for reasoning, coding, and writing when compared with leading AI models from competitors like Google, OpenAI, and Anthropic.
People familiar with the matter told The New York Times that the model, code-named Avocado, outperformed Meta’s previous model and Gemini 2.5, which launched last March. However, it did not perform as well as Gemini 3.0, which was released in November.
Meta is now reportedly delaying the rollout of Avocado until at least May. The model had originally been planned for release in March. Meta’s AI team has discussed temporarily using Google’s Gemini models for its products.
“As we’ve said publicly, our next model will be good, but more importantly, show the rapid trajectory we’re on, and then we’ll steadily push the frontier over the course of the year as we continue to release new models,” a Meta spokesperson told Gizmodo in an emailed statement. “We’re excited for people to see what we’ve been cooking very soon.”
The statement closely followed what CEO Mark Zuckerberg told investors on a call in January.
The report comes as Meta, the parent company of Facebook, Instagram, and WhatsApp, has appeared laser-focused on AI. Meta has committed to spending $600 billion on AI infrastructure in the U.S. by 2028. The company also projects its 2026 capital expenditures to reach between $115 billion and $135 billion, this year alone. Zuckerberg has said the company is ultimately trying to build artificial “superintelligence,” a type of AI that could exceed human intelligence.
It has also splurged on large pay packages, some reportedly in the $100 million range, to lure talent from other AI companies.
Andrew Tulloch, co-founder of Thinking Machines Lab, joined Meta in October. Some reports claim his pay package could be worth up to $1.5 billion over at least six years, tied to performance bonuses and stock incentives.
In June, Meta bought a 49% stake in Scale AI, with the startup’s founder, Alexander Wang, joining Meta as its Chief AI Officer. The New York Times reports that Wang assembled a team called the TBD Lab within the company’s AI division that was working on new models, including Avocado. There have also been reports that Wang has clashed with other Meta executives, including Zuckerberg himself. A Meta spokesperson posted on X that those rumors were “Totally false.” Earlier this week, Zuckerberg posted a selfie with Wang on Threads with the caption “Meanwhile at Meta HQ.”
Source: Gizmodo