Meta, the parent company of Facebook, Instagram, and Threads, is delaying the release of its next-generation artificial intelligence model, dubbed “Avocado,” after internal testing revealed performance shortcomings compared to rival systems. The delay, reported by The New York Times, comes as Meta explores the possibility of temporarily licensing Google’s Gemini AI to power its products, signaling a potential shift in strategy for the tech giant.
The setback represents a significant challenge for Meta CEO Mark Zuckerberg, who has invested heavily in AI development, committing an estimated $600 billion to building data centers and spending as much as $135 billion this year alone on AI initiatives – nearly double the $72 billion spent last year, according to company projections made in January. Despite these substantial investments, the “Avocado” model reportedly fell short in key areas like reasoning, coding, and writing during internal evaluations.
Originally slated for release this month, the launch of Avocado is now expected around May. While the model outperformed Meta’s previous AI iteration and even surpassed Google’s Gemini 2.5 (released in March), it lagged behind the more recent Gemini 3.0, launched in November. Zuckerberg stated in a January investor call, “I expect our first models will be good, but more importantly will show the rapid trajectory we’re on,” highlighting the pressure to deliver competitive AI capabilities.
The potential licensing of Google’s Gemini underscores the competitive landscape of the AI industry, where Google, OpenAI, and Anthropic are currently considered leaders in foundational AI models. These models underpin technologies like chatbots, video generators, and coding tools. Meta’s AI division, led by Chief AI Officer Alexandr Wang, has been working to close the gap after its previous model, Llama 4, did not meet expectations.
Internal Tensions and Strategic Shifts
The delay and potential licensing agreement also reveal internal discussions and strategic adjustments within Meta. The Times reported disagreements between Wang and other top executives, including Chief Product Officer Chris Cox and Chief Technology Officer Andrew Bosworth, regarding how new AI models should contribute to Meta’s advertising business. Last week, Meta announced the creation of an AI team under Bosworth to collaborate with Wang, as reported by The Wall Street Journal.
Wang’s AI lab, known as TBD Lab, completed the initial “pre-training” phase of Avocado development at the end of last year and began “post-training” in January. The lab, comprised of around 100 employees, has reportedly experienced some staff turnover prior to the anticipated launch. Vibes, an AI video app similar to OpenAI’s Sora, is currently the only product released by Wang’s division.
Open Source Debate and Investment in AI Talent
Meta has historically advocated for open-source AI models, believing they foster technological advancement. However, companies like OpenAI and Anthropic have expressed concerns that making AI code publicly available could pose safety risks. The debate over whether to release Avocado as an open-source model continues internally.
To accelerate its AI development, Meta invested $14.3 billion in Scale AI last June and appointed Wang, the 29-year-old CEO of Scale AI, as its Chief AI Officer. This move signaled Zuckerberg’s commitment to assembling a top-tier AI team and rapidly advancing Meta’s capabilities in the field. The company’s substantial financial commitment to AI infrastructure reflects the scale of its ambition.
What’s Next for Meta’s AI Strategy?
The delay of Avocado and the consideration of licensing Google’s Gemini represent a pivotal moment for Meta’s AI strategy. The company’s next steps will likely determine its position in the rapidly evolving AI landscape. Meta maintains it will “steadily push the frontier” with new models, according to a company spokesperson. The coming months will reveal whether Meta can overcome the current challenges and deliver AI products that meet its ambitious goals.
The situation highlights the intense competition and technical complexities inherent in developing cutting-edge AI. As Meta navigates these challenges, its decisions will have significant implications for the future of its products and its role in the broader technology industry. Share your thoughts on Meta’s AI strategy in the comments below.
