Meta has postponed the release of its text-based large language model, named Avocado, until at least May, according to the New York Times.
The AI model, which had been slated for launch this month, is reportedly falling short of performance benchmarks set by Google’s Gemini 2.5 and Gemini 3.
The delay comes despite Meta’s heavy investment in AI, including plans to build its own chips. In January, the company announced capital spending of between $115 billion and $135 billion for the year, part of its pursuit of “superintelligence”, the point at which AI surpasses human capability.
“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 Reuters, resonating with the remarks made by CEO Mark Zuckerberg during an earnings call earlier this year.
The spokesperson added, “We’re excited for people to see what we’ve been cooking very soon.”
The NYT report also suggested that Meta’s AI division has discussed temporarily licensing Google’s Gemini to power its products, though no decision has been made.
Media outlets previously reported in December that Meta was working on Avocado, a text‑based AI model originally scheduled for a first‑quarter launch.
According to a previous WSJ report, Meta is also building “Mango” as a high-stakes image- and video-focused AI model.
ALSO READ: NTT DATA Furthers AI Deployment at Scale with NVIDIA-Powered Enterprise AI Factories
