Meta Delays 'Avocado' AI Model Release After Falling Short of Rivals

Meta Platforms is pushing back the launch of a highly anticipated artificial intelligence model after internal testing revealed it couldn't keep pace with competing systems from Google and other leading tech companies. The model, internally known as "Avocado," has been delayed until at least May 2026, according to a New York Times report citing people familiar with the situation.

Although Avocado demonstrated stronger performance compared to Meta's previous AI models, it failed to match the capabilities of Google's Gemini 3.0, which launched in November. This performance gap prompted Meta's leadership to hold off on a public release rather than debut a product that trails the current industry standard.

To bridge the gap in the short term, Meta is reportedly weighing the option of licensing Google's Gemini technology to keep its AI-powered features competitive while Avocado undergoes further development. This would mark a notable shift for a company that has consistently positioned itself as a developer of its own foundational AI technology.

The delay comes despite Meta's massive and ongoing investment in artificial intelligence infrastructure. Over the past several years, the company has poured hundreds of billions of dollars into AI research, aggressively recruited top-tier talent from across the industry, and expanded its computing resources to support large-scale model training. Earlier this year, Meta indicated it plans to allocate up to $135 billion toward AI development throughout 2025.

The setback highlights just how competitive the AI landscape has become, with Google, OpenAI, and other players continuously raising the bar. For Meta, catching up — or pulling ahead — will require not just capital, but consistent execution at a level the Avocado delay suggests still needs refinement.