Key Takeaways
- Meta’s upcoming AI model, internally called “Avocado,” won’t arrive until May 2026 at the earliest.
- Internal benchmarks show Avocado’s capabilities sit somewhere between Google’s Gemini 2.5 and Gemini 3.
- The social media giant is exploring a potential licensing agreement with Google for Gemini AI technology.
- Meta’s previous flagship AI release — Llama 3 — debuted in 2024.
- This postponement occurs while Meta projects capital spending of up to $135 billion throughout 2026.
Meta Platforms has postponed its highly anticipated “Avocado” AI model until May following disappointing internal performance reviews. Sources indicate the company is exploring a short-term licensing arrangement with Google Gemini.
Why Meta’s AI Release Is Being Pushed Back
Meta Platforms has rescheduled the debut of its upcoming AI system, known internally as “Avocado,” moving the timeline from April to at least May — with a potential June launch also under consideration.
The New York Times initially broke the story, drawing on insider information. Reuters independently verified that Avocado’s current capabilities land somewhere between Google’s Gemini 2.5 and Gemini 3 models — suggesting performance gaps remain.
Meta’s most recent frontier AI model launch was Llama 3 in 2024. In the interim, competitors including Google’s Gemini, OpenAI’s GPT series, and Anthropic’s Claude have advanced significantly.
The company did introduce two scaled-down Llama 4 variants last year, optimized for resource-constrained environments. These were meant to precede a larger, more capable system dubbed “Behemoth” — though updates on that project have gone quiet.
A Meta representative stated: “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 continued: “We’re excited for people to see what we’ve been cooking very soon.”
Meta stock declined approximately 2% during Friday’s opening hours.
Is a Google Partnership on the Table for Meta?
The situation takes an intriguing turn here. Senior figures within Meta’s artificial intelligence teams have reportedly explored temporarily licensing Google’s Gemini technology to support company products during Avocado’s development. While no final determination has been reached, these discussions are actively underway.
Such a move would represent a significant concession for CEO Mark Zuckerberg. He’s committed substantial resources to his Superintelligence Labs initiative, assembling elite AI researchers and engineers. Meta’s projected capital expenditures for 2026 range from $115 billion to $135 billion, predominantly allocated toward AI-related infrastructure.
Licensing Gemini would essentially provide Alphabet with a high-profile victory — transforming its image from AI follower to a technology provider that competitors rely upon.
Meta has already generated tangible AI-driven value — enhanced advertising precision, superior content curation, and proprietary silicon for recommendation systems. The company also leads in the smart glasses category.
However, Zuckerberg has openly pledged to develop a “personal superintelligence” — a Meta-created platform potentially capable of rivaling Apple and similar smartphone ecosystems as consumers’ primary computing interface. Achieving this vision demands proprietary, industry-leading AI models rather than licensed alternatives.
Currently, the Avocado postponement dominates the narrative. Meta stock traded down roughly 2% in early Friday sessions.



