Meta Avocado AI Model Delayed: Why It’s Not Ready and What Comes Next

Meta’s next flagship AI model, codenamed Avocado, is not ready.

The company planned to release it this month. It will not ship until May 2026 at the earliest, and possibly June.

Meta’s Avocado AI model was delayed because it wasn’t good enough to compete with its rivals in internal testing.

Specifically, when Meta ran benchmarks, Avocado beat their previous models and Google’s Gemini 2.5, but it couldn’t match Gemini 3.0. It also fell short of the top models from OpenAI and Anthropic.

For a company about to spend $115-135 billion on AI in 2026 and publicly promising to reach superintelligence, releasing a model that sits behind the current frontier wasn’t acceptable. So Meta pushed the launch from March to May or June 2026 to keep improving it.

That’s the whole reason. The model wasn’t ready. The benchmark gap with Gemini 3.0 was too wide to ship.

This is the latest setback in Meta’s effort to compete at the frontier of AI, and it is happening while the company is spending up to $135 billion on AI this year alone.

What is Meta’s Avocado AI Model?

Avocado is a next-generation text large language model being developed inside Meta Superintelligence Labs, aimed at significantly improving coding and reasoning capabilities over Meta’s current Llama-based systems.

A next-gen follow-up model codenamed Watermelon is already planned. Meta is also building an image and video generator codenamed Mango.

All three projects are being developed under Alexandr Wang, the Scale AI co-founder who joined Meta as chief AI officer. His division, internally called TBD Lab, is tasked with building what Zuckerberg has described as superintelligent AI systems.

Why Has Meta Delayed the Avocado AI Launch?

In internal tests, Avocado fell short of leading models from Google, OpenAI, and Anthropic in logical reasoning, programming, and writing. It beat Meta’s previous model and Google’s Gemini 2.5 but could not match Gemini 3.0.

According to a person familiar with the situation, the model’s performance currently sits somewhere between Gemini 2.5 and Gemini 3 from Google. The gap with rival systems prompted Meta to delay the rollout until May or possibly June while improvements are made.

In an internal memo from early February 2026, Meta product manager Megan Fu described Avocado as “Meta’s most capable pre-trained base model to date” and said it was “competitive with leading post-trained models even before undergoing post-training refinement.” That optimism has not translated into a launch-ready product.

A Meta spokesperson told Reuters: “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.”

This is the second consecutive missed deadline. Meta insiders initially expected the model before the end of 2025, but it slipped to the first quarter of 2026 amid training and performance-testing challenges.

Meta Is Considering Licensing Google’s Gemini Technology

This is the detail that stands out most.

Meta’s leadership discussed temporarily licensing Gemini to power its AI products while Avocado development continues, though no decision was made.

For a company spending $115-135 billion on its own AI infrastructure this year, including a roadmap for building its own chips, the prospect of licensing a competitor’s model to fill the gap is a significant signal about where Avocado stands right now.

How Does Avocado Compare to Rivals?

Meta’s Llama models were well-received by the developer community in earlier generations. Llama 4’s April release failed to captivate developers, triggering what sources describe as a major internal shake-up. Avocado was supposed to be the answer. The delay means Meta remains behind its main competitors at the frontier, for now.

What This Means for Meta’s $135 Billion AI Bet

Meta is not struggling financially. The problem is competitive positioning.

In January, Meta laid out capital-spending plans of between $115 billion and $135 billion for the year in the pursuit of “superintelligence,” the horizon where AI will outsmart humans.

Meta’s stock has underperformed the broader tech sector this year while trailing Google parent Alphabet. Wall Street wants to see returns on the company’s raised capital expenditure guidance.

Zuckerberg has told investors he believes Meta has built “the lab with the highest talent density in the industry.” The Avocado delay tests that claim directly. Alexandr Wang arrived with a mandate to close the gap with OpenAI and Google. So far, that gap has not closed.

Key Facts: Meta Avocado AI Delay

• Original launch target: March 2026 (Q1)
• New launch target: May–June 2026 at earliest
• Reason: Failed to match Gemini 3.0 in reasoning, coding, and writing benchmarks
• Performance level: Between Gemini 2.5 and Gemini 3.0
• Backup option being considered: Temporarily licensing Google Gemini
• Meta’s 2026 AI spend: $115–135 billion
• Division responsible: Meta Superintelligence Labs (TBD Lab), led by Alexandr Wang.
• Next models in pipeline: Mango (image/video), Watermelon (next-gen text).

FAQ: Meta Avocado AI Model

Why Was Meta Avocado AI Delayed?

Avocado did not perform well enough in internal benchmarks. It beat Meta’s previous models and Gemini 2.5 but could not match Gemini 3.0 from Google. Meta pushed the launch to May or June 2026 to allow more time for improvement.

When Will Meta Release the Avocado AI Model?

The current target is May 2026 at the earliest. It could slip to June. This is the second delay, it was originally planned for late 2025, then Q1 2026, and now Q2 2026.

How Does Avocado Compare to ChatGPT and Gemini?

Based on internal testing, Avocado currently falls between Google’s Gemini 2.5 and Gemini 3.0. It has not yet matched the top tier of models from OpenAI, Google, or Anthropic.

Is Meta Licensing Google Gemini?

Meta has reportedly discussed temporarily licensing Google’s Gemini technology to power its AI products while Avocado development continues. No final decision has been confirmed.

What is Meta Spending on AI in 2026?

Meta announced capital expenditure plans of $115-135 billion for 2026, directed at AI infrastructure, model development, and building proprietary chips.

Conclusion

Meta Avocado AI is delayed because it is not competitive enough, not yet. The company has the money, the infrastructure, and now a dedicated frontier AI division under Alexandr Wang. What it does not have is a model that can match Gemini 3.0 or GPT-5 at release.

The delay is a setback, but not a collapse. Meta is spending at a scale that few companies can match. The question is whether May or June delivers a model that actually closes the gap with Google and OpenAI, or whether Avocado arrives as another near-miss.