Falling behind its rivals and facing internal tension, Meta reportedly preps new “Avocado” AI model
2025 turned out to be quite a chaotic year for Meta’s big AI dreams.
This year was supposed to be all about Llama 4, Meta’s open-source AI model that never fully launched.
In a frenzied pivot to get back in the race, Mark Zuckerberg undertook an AI all-star hiring spree for his new Meta Superintelligence Labs, spending oodles of cash on NBA-sized pay packages to lure recruits.
So how’s it all going within the company? Things aren’t exactly humming along, according to CNBC.
It reports that the new team has encountered friction within Meta’s corporate structure. The cloistered team is apparently working on a new frontier AI model code-named “Avocado,” which, despite Mark Zuckerberg’s passionate open-source AI manifesto, might turn out to be a proprietary, closed-source model.
Per the report, the many in the company were expecting Avocado to be released before the end of this year, but it’s now planned for Q1 2026.
A Meta spokesperson said, “Our model training efforts are going according to plan and have had no meaningful timing changes.”
Updated to include comments from Meta
In a frenzied pivot to get back in the race, Mark Zuckerberg undertook an AI all-star hiring spree for his new Meta Superintelligence Labs, spending oodles of cash on NBA-sized pay packages to lure recruits.
So how’s it all going within the company? Things aren’t exactly humming along, according to CNBC.
It reports that the new team has encountered friction within Meta’s corporate structure. The cloistered team is apparently working on a new frontier AI model code-named “Avocado,” which, despite Mark Zuckerberg’s passionate open-source AI manifesto, might turn out to be a proprietary, closed-source model.
Per the report, the many in the company were expecting Avocado to be released before the end of this year, but it’s now planned for Q1 2026.
A Meta spokesperson said, “Our model training efforts are going according to plan and have had no meaningful timing changes.”
Updated to include comments from Meta
