Meta has officially postponed the highly anticipated release of its new artificial intelligence model, codenamed Avocado. Initially scheduled to launch at the start of this year, the rollout has been pushed back to at least May following disappointing results during internal testing. The delay stems from the model’s inability to match the performance benchmarks set by competing systems from major industry rivals.
According to a report by the New York Times, the new system failed to outperform the latest large language models currently dominating the market. As Meta delays AI model Avocado, the company faces mounting pressure to deliver top-tier technology while juggling massive infrastructure investments and potential sweeping workforce reductions across its global operations.
Falling Behind Competitor Benchmarks
During Meta’s internal evaluations, the new text-based system struggled to surpass the capabilities of Google’s Gemini 3, a highly capable model that debuted in November of last year. However, the tests did show that the delayed system managed to achieve notably higher scores than both Google’s older Gemini 2.5 and Meta’s own previous generation system, Llama 4.
Despite showing clear improvement over these older models, falling behind the latest offerings from OpenAI and Anthropic prompted the company to pause the public release. Rather than releasing a product that trails behind the current industry standard, Meta leadership has chosen to take additional time to refine the system before bringing it to millions of global users.
Advancing Autonomous Agentic Technology
Avocado represents the first major project emerging from Meta’s newly established Superintelligence Labs, known internally as MSL. This dedicated research division is spearheaded by Alexandr Wang. Meta demonstrated its commitment to this leadership by acquiring Wang’s previous startup, Scale AI, for $14.5 billion last year. The laboratory is specifically focused on pioneering next-generation capabilities that go beyond simple chat interfaces.
Internal documents outline Avocado as a highly specialized system carefully optimized for complex technical tasks. It is explicitly designed to handle software development, logical reasoning, and sophisticated “agentic” behavior. The primary goal is for the system to function as an autonomous agent that can independently handle multi-step planning and execute intricate digital workflows without requiring constant human input.
To further strengthen these agentic capabilities, Meta recently acquired Moltbook, a unique and viral social media platform where only artificial intelligence bots are permitted to interact. The platform famously gained attention when its bots began demanding a takeover of the world. The entire Moltbook team will integrate directly into the Superintelligence Labs division to accelerate the development of autonomous systems. Additionally, the MSL division is working on a separate image-generation tool codenamed Mango.
Exploring Temporary Licensing Solutions
The current delay leaves a noticeable temporary gap in Meta’s immediate product roadmap. The company currently powers its dedicated chatbot application and various smart features across Facebook, Instagram, and WhatsApp using its own proprietary technology. To prevent active users from migrating to competitor platforms during this delay, company executives are reportedly considering alternative solutions.
One heavily discussed strategy involves temporarily licensing Google’s Gemini technology for use across Meta’s massive ecosystem of social applications. While no final agreements have been formalized, bringing in an external system could help maintain user engagement and provide advanced features while internal development continues on the delayed Avocado project.
Massive Costs and Sweeping Layoffs
Developing cutting-edge technology requires immense financial resources. Meta plans to invest between $115 billion and $135 billion on its technical framework and data centers this year alone. These soaring capital expenditures highlight the massive cost of staying relevant in the current tech landscape.
To offset these rising infrastructure costs, the company is preparing for a massive internal restructuring effort. According to an exclusive report from Reuters, Meta is planning a sweeping layoff that could eliminate up to 20 percent of its total global workforce. This aggressive downsizing is intended to balance the budget and prepare the organization for a future where a highly efficient, AI-assisted workforce requires significantly fewer human employees.
Top executives have already communicated the potential downsizing plan to senior leaders across various departments. Management has been directly instructed to begin formulating targeted strategies for staff reductions. As the company navigates these expensive technical hurdles and broad internal changes, the tech industry will be closely watching the rescheduled May launch date to see if the massive investments will pay off.
