Meta delays launch of AI model Avocado in latest setback

Meta The last fruit of $META’s AI is ripening on company time. The New York Times reported Thursday that Meta pushed back Avocado – the text model intended to headline CEO Mark Zuckerberg’s latest and greatest AI reset – from a planned March debut to at least May after internal testing showed it was lagging behind major AI systems. Google $GOOGL, OpenAI and Anthropic on basically everything: reasoning, coding and writing.

It’s another horrible moment for a company that spent lavishly to convince investors, employees and Silicon Valley that it could claw its way back to the forefront of AI. In January, Zuckerberg told investors that MetaThe first new models would be “good” and show the “rapid trajectory” the company is on. The same month, the company said it planned capital spending of $115 billion to $135 billion for 2026, with the increase due in part to Meta Superintelligence Laboratories. But writing giant checks and building giant data centers still doesn’t guarantee a place at the front of the model race.

The Times, citing unnamed sources familiar with the matter, reported that Avocado currently lands somewhere between GoogleThese are Gemini 2.5 and Gemini 3. It’s hard to come back to after months and months of preparation, because it doesn’t suggest any real progress on the one thing this project was supposed to do: get people in AI talking about the company as peers again.

Building a boundary model is difficult. Shipping one that can actually stand next to the best systems GoogleOpenAI and Anthropic is more difficult. MetaThe problem is that he has made the stakes of this challenge impossible to ignore. Meta needs a model that changes the conversation around business. The lawyer, at least for now, seems to be doing the opposite.

It’s not MetaThis is the first awkward moment in the lab. Llama 4 was delayed after failing to meet MetaGoogle’s expectations for mathematical reasoning and credentials even after Zuckerberg said – in January 2025 – that he expected it to “become the leading edge model” that year. The model was not particularly well received. In May, Meta delayed its flagship Behemoth model because engineers struggled to produce significant enough improvements. By the end of June, Zuckerberg had reorganized the company’s AI work under Meta Superintelligence Laboratories. NOW, MetaAvocado leaders are already thinking beyond Avocado. The next model, the Times reported, is called Watermelon.

Leaders within MetaAI division discussed temporary license GoogleGemini will power some of the company’s AI products while Avocado catches up, the Times reports. When the company spends like it wants to own the next IT platform, it would discuss leasing GoogleIn the meantime, the catch-up story is starting to look even more expensive than its bills already are.

Meta already tried the expensive reset. This was supposed to be the quarter where the company began to show that all the spending bills, all the futuristic promises, all the talent drama, and all the superintelligence rhetoric were producing something more compelling than ambition. Instead, the image now seems uncomfortably familiar. And Avocado feels like another reminder that throwing billions of dollars at a problem doesn’t make the credentials blink.

Meta can further bridge the AI ​​gap. Rich corporations get more of a hit, and Zuckerberg clearly intends to keep paying for the bats. The company has extensive distribution and extensive infrastructure projects. But the public scorecard currently shows a company that has spent heavily, hired aggressively, reorganized repeatedly, and has yet to explain why its next big thing is late — and why one of the options on the table relied on the rival it is ultimately supposed to beat.

Investors were ready to indulge MetaAI’s appetite for AI because its core business continues to print money. The advertising machine remains a monster, and that cash flow has helped make the company’s AI madness, if not exactly calm, at least defensible. When Meta released its fourth-quarter results in January, shareholders rewarded the company even as it presented a spending plan that would have seemed deranged almost anywhere else. Zuckerberg called 2026 a big year for achieving personal superintelligence. This type of line works much better when the next major model arrives on time and looks like a real leap forward.

But the race for the frontier model is different. This is the prestige competition. It’s here GoogleOpenAI and Anthropic look like adults, and where Meta has spent the last year trying to prove that he shouldn’t sit at the kids’ table anymore. A delayed flagship model doesn’t really help make that point.

For now, Avocado looks like another reminder that frontier AI has been a humbling affair for Meta. The company has money, compute, distribution, and one of the most determined CEOs in tech. What Meta what’s still missing is the clean, undeniable launch that makes all this spending seem like forethought rather than impatience.