TLDR
- Meta has postponed its next major AI model, codenamed “Avocado,” until at least May 2026.
- The model’s performance currently lies between Google’s Gemini 2.5 and Gemini 3.
- Meta is reportedly considering temporarily licensing Google’s Gemini to power its AI products.
- Meta last released a cutting-edge AI model, Llama 3, back in 2024.
- The delay occurs as Meta plans to spend up to $135 billion on capital expenditure in 2026.
Meta’s “Avocado” AI model has been delayed until at least May after internal testing revealed it was inferior to competitors. The company is reportedly considering a temporary license deal with Google’s Gemini.
What’s Going On With Meta’s Next AI Model?
Meta Platforms has delayed the release of its next AI model, internally codenamed “Avocado,” pushing it back from this month to at least May – and potentially as late as June.
Meta Platforms, Inc., META
The New York Times first reported the delay, citing sources familiar with the matter. Reuters separately confirmed that the model’s current performance falls between Google’s Gemini 2.5 and Gemini 3 – that is, it’s not where Meta wants it to be.
The last time Meta released a truly cutting-edge AI model was Llama 3 in 2024. Since then, Google’s Gemini, OpenAI’s GPT, and Anthropic’s Claude have all advanced.
Meta did release two lighter versions of Llama 4 last year, designed to run on limited hardware. Those were supposed to be followed by a larger, more powerful model called “Behemoth” – but there’s been no news on that front either.
A Meta spokesperson said: “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.” The spokesperson added: “We’re excited for people to see what we’ve been working on very soon.”
Meta stock dropped around 2% in early Friday trading.
Could Meta End Up Licensing Google’s AI?
Here’s where it gets interesting. Leaders within Meta’s AI division have reportedly discussed temporarily licensing Google’s Gemini to power the company’s products while Avocado is being completed. No decision has been made yet, but the discussion is taking place.
That would be a hard thing for CEO Mark Zuckerberg to accept. He has invested heavily in his Superintelligence Labs division, gathering some of the industry’s top AI talent. The company has laid out capital spending plans of between $115 billion and $135 billion for 2026, much of it aimed at building AI infrastructure.
A Gemini licensing deal would effectively give Alphabet a public victory – changing its reputation from being seen as an AI laggard to a company others are now paying to use.
Meta has already achieved real returns from AI – better ad targeting, improved content recommendations, and its own in-house chips for recommendation models. It also has a dominant position in the smart glasses market.
But Zuckerberg has publicly committed to building a “personal superintelligence” – a Meta-developed system that could eventually challenge Apple and other smartphone platforms as the preferred everyday computing device. That goal requires Meta to have a leading model, not a licensed one.
For now, the Avocado delay remains the main story. Meta stock was down around 2% on Friday as of early trading.
