Meta’s high-profile AI team is facing mounting pressure as the company pushes to rebound from the disappointing rollout of Llama 4, which drew criticism for weak real-world performance, limited coding capabilities, and disputed benchmark scores. The urgency has intensified as Meta works on “Avocado,” a next-generation frontier model being developed inside its internal TBD Lab.
According to individuals familiar with the matter, senior AI leader Wang has been tasked with delivering a top-tier model capable of restoring Meta’s competitiveness against rivals. The pressure has escalated following strong industry responses to Google’s Gemini 3.0, OpenAI’s GPT-5 and GPT-5.2, and Anthropic’s Claude Opus 4.5, all of which have set new expectations for the AI landscape.
Meta’s top AI executive and former GitHub CEO Nat Friedman is also under scrutiny. He previously oversaw the launch of Vibes, an AI-driven short-video feed unveiled in September. However, insiders say the product has been widely seen as inferior to OpenAI’s Sora 2, partly due to limited features and rushed development.
Inside the company, the pressure is affecting employees as well. Workers are reportedly clocking 70-hour weeks following Meta’s extensive restructuring and job cuts over the past year. Many employees are concerned about the shift away from Meta’s traditional collaborative approach, which involved multiple teams managing design, privacy, algorithms, and user experience.
Sources note that current processes are not well-suited for building AI foundation models. Friedman is pushing for newer, streamlined tools built for multi-model integration and advanced coding automation, arguing that faster development cycles are essential for Meta to keep pace.
As competitors continue to raise the bar, Meta’s ability to deliver an industry-leading AI breakthrough will be critical in determining its future position in the global AI race.
Follow Startupbydoc for daily startup insights, funding news, IPO analysis, and business breakdowns.