Why Meta Is Switching to Paid AI Models After Years of Free Access
- Meta is abandoning its long-standing "open-source" strategy to build a closed, money-making AI model codenamed "Avocado" after the failure of its Llama 4 release.
- CEO Mark Zuckerberg has hired Alexandr Wang in a $14.3 billion deal and pushed out veteran staff like Yann LeCun while instituting a high-pressure culture with 70-hour workweeks.
- The company is reportedly training its new AI using data from rivals including Chinese firms, which contradicts Zuckerberg's previous political stance against Chinese technology.
Meta Platforms is currently undergoing one of the most significant changes in its history as CEO Mark Zuckerberg steers the company in a new direction. After years of championing "open source" artificial intelligence where the code is available for anyone to use, Meta is now pivoting toward building money making products that it keeps under lock and key. This shift is driven by the development of a new AI model codenamed "Avocado" which is expected to launch next spring. Unlike previous models that were free for developers to tinker with, Avocado will likely be a "closed" system designed to compete directly with paid services from Google and OpenAI.
The decision to change course comes after a major disappointment earlier this year involving Meta’s previous flagship model known as Llama 4. The release failed to impress the tech industry and left Zuckerberg frustrated with the company's progress. In response he has taken a much more hands on approach to daily operations. He reorganized the AI division and sidelined long time employees while bringing in new blood from outside the company. This includes the high profile hiring of Alexandr Wang, a 28 year old tech prodigy whose company Scale AI received a massive $14.3 billion investment deal from Meta.
Zuckerberg is sparing no expense to win the AI race and has pledged to spend up to $600 billion on infrastructure over the next three years. However this spending spree has caused tension on Wall Street where investors are worried about the rising costs and the lack of immediate profit. Meta’s stock has underperformed compared to rivals like Google, and shareholders are anxious to see a return on investment. The company hopes that moving away from free open source software to profitable proprietary models will calm these financial fears.
The development of the new Avocado model involves some controversial methods. Sources say that Meta is training its new AI by using technology from its biggest rivals, including Google and OpenAI. Perhaps most surprisingly the team is also using models from Chinese tech giants like Alibaba. This contradicts Zuckerberg's previous warnings about the dangers of Chinese AI technology and state censorship. It suggests that in the desperate race to catch up Meta is willing to use whatever tools are available regardless of their origin.
Inside the company the culture is shifting rapidly from a relaxed academic environment to a high pressure startup vibe. Reports indicate that 70 hour workweeks have become the new normal for the AI teams. The new "TBD Lab" group led by Wang is situated right around Zuckerberg’s desk to allow for constant supervision. This micromanagement style has alienated some veteran staff including Yann LeCun, one of the "godfathers of AI," who recently left the company due to disagreements over resources and strategy.
Meta is also struggling to release successful consumer products that utilize this new technology. To beat OpenAI to the market Meta rushed out a video generation tool called "Vibes" in September. However the product received mixed reviews and was quickly overshadowed when OpenAI released its superior Sora 2 model just a week later. These product stumbles have increased the pressure on the new leadership team to deliver a hit with the upcoming Avocado model.
The terminology used by the company is also causing some confusion and concern. Zuckerberg has stated that the goal is to achieve "superintelligence" which refers to computers that are smarter than humans. Market research conducted by the company found that this term scares many people and regulators, especially in Europe. Despite the public unease Zuckerberg remains committed to this vision and believes that building the most advanced AI is the only way for Meta to stay relevant in the future.
As 2026 approaches Meta finds itself in a precarious position. It is trying to reinvent its business model while fighting off fierce competition and managing internal turmoil. The company is betting its future on Alexandr Wang and the new closed model strategy. If Avocado succeeds it could validate Zuckerberg's expensive pivot. If it fails Meta risks falling further behind in a technology race that it can no longer afford to lose.