Meta Delays Rollout of New A.I. Model After Performance Concerns

The tech giant pushed back the timeline after spending billions to be on the cutting edge of artificial intelligence.

Mar. 12, 2026 at 11:36pm

Got story updates? Submit your updates here. ›

Meta has delayed the release of its new 'frontier' A.I. model, code-named Avocado, to at least May from this month due to performance concerns. The model has fallen short of the performance of leading A.I. models from rivals like Google, OpenAI and Anthropic on internal tests for reasoning, coding and writing. As a result, Meta has considered temporarily licensing Google's Gemini model to power its A.I. products.

Why it matters

Meta has staked the future of the company on being at the forefront of A.I. development, investing billions to hire top researchers and build data centers. The performance of its new A.I. model is being closely watched in the competitive race for advanced AI capabilities that can power new products and services.

The details

Meta's new 'frontier' A.I. model, Avocado, outperformed the company's previous model and did better than Google's Gemini 2.5 model from March, but it has not performed as strongly as Gemini 3.0 from November. As a result, Meta has delayed Avocado's release to at least May from this month. The leaders of Meta's A.I. division have discussed temporarily licensing Gemini to power the company's A.I. products, though no decisions have been reached.

  • Meta finished the first stage of Avocado's development, called 'pre-training', at the end of last year.
  • In January, Meta began the next phase, 'post-training', and set a target release date of mid-March for Avocado.

The players

Mark Zuckerberg

The chief executive of Meta, who has invested billions in the A.I. race and staked the future of the company on being at the cutting edge of A.I. development.

Alexandr Wang

The 29-year-old chief executive of Scale AI, whom Meta hired as its new chief A.I. officer after investing $14.3 billion in the start-up.

Chris Cox

Meta's chief product officer, who has clashed with Alexandr Wang over how the new A.I. models should improve the company's advertising business.

Andrew Bosworth

Meta's chief technology officer, who will be leading a new A.I. engineering team that will collaborate with Alexandr Wang and the A.I. division.

Dave Arnold

A spokesman for Meta.

Got photos? Submit your photos here. ›

What they’re saying

“I expect our first models will be good, but more importantly will show the rapid trajectory we're on.”

— Mark Zuckerberg (The New York Times)

“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. We're excited for people to see what we've been cooking very soon.”

— Dave Arnold, Spokesman, Meta (The New York Times)

What’s next

The leaders of Meta's A.I. division have discussed temporarily licensing Google's Gemini model to power the company's A.I. products, though no decisions have been reached.

The takeaway

Meta's struggles to develop a cutting-edge A.I. model that can compete with industry leaders like Google, OpenAI, and Anthropic highlight the challenges of being at the forefront of rapidly evolving AI technology. The delay in releasing Avocado raises questions about Meta's ability to deliver on its ambitious AI goals and the company's long-term strategy in the AI race.