Meta delays release of new AI, weighs licensing Google’s Gemini after disappointing trial runs: report
Meta Delays Release of Next AI Model Amidst Competition
Meta, led by Mark Zuckerberg, has decided to postpone the release of its upcoming artificial intelligence model, “Avocado,” by at least two months. Internal tests revealed that the technology was not up to par with competitors like Google, OpenAI, and Anthropic.
Considering the underwhelming performance of Avocado, Meta’s AI division is contemplating utilizing Google’s Gemini temporarily to power its products, as reported by The New York Times.
Despite months of work by Meta’s tech team, Avocado did not meet the expected standards in reasoning, coding, and writing during internal testing.
Originally scheduled for release this month, Avocado is now expected to launch around May, as it fell short in various aspects compared to its competitors.
Mark Zuckerberg had expressed optimism in July about Meta’s AI advancements, stating that the new models would push boundaries in the following year.
While Avocado showed improvement over Meta’s previous AI model and Google’s Gemini 2.5 from March, it still lagged behind Gemini 3.0 from November.
A Meta spokesperson reassured that they will continue to innovate and release new models throughout the year, expressing excitement about the upcoming developments.
Meta’s AI endeavors are crucial in the ongoing artificial intelligence race, with Google, OpenAI, and Anthropic leading the way in foundational AI models for various applications.
Meta’s Intensified AI Efforts
Following the underperformance of Meta’s previous AI model, Llama 4, the company invested significantly in the startup Scale AI and appointed Alexandr Wang as the new Chief AI Officer.
Wang spearheaded the establishment of an AI lab within Meta called TBD Lab, where the development stages of Avocado took place. The lab recently completed the pre-training stage and began post-training in January.
Despite setbacks, Meta released Vibes, an AI video app similar to OpenAI’s Sora, under Wang’s leadership.
There have been some departures within TBD Lab before Avocado’s release, with reports of turnover among researchers.
Debates have arisen within Meta about whether the new AI model should be open-source, potentially allowing other developers access to its code.
Meta’s advocacy for open-source models contrasts with the concerns of competitors like OpenAI and Anthropic, who believe sharing AI code could pose safety risks.
Tensions between Wang and other Meta executives, Chris Cox and Andrew Bosworth, have been noted, particularly regarding the impact of new AI models on Meta’s advertising business.
Meta recently announced the formation of an AI team under Bosworth that will collaborate with Wang to further AI development.
