Meta shifts from an open-source strategy to paid AI models, starting with the new Avocado model.

Meta changes development course: Zuckerberg bets on closed AI models

The company Meta is abandoning open artificial intelligence models in favor of closed solutions, preparing to launch a new product under the code name Avocado. This is part of a new strategy aimed at commercial gain.

Meta is preparing to launch the new AI model Avocado, which is scheduled for spring 2026. Unlike the previous Llama model, which was available to researchers, Avocado will become a “closed” model, allowing for commercial access. This step indicates that Meta is starting to adopt a commercial strategy similar to OpenAI and Google to profit from its innovations in the field of artificial intelligence.

The move away from open models resulted from disappointment with the results of Llama 4, the development of the next version of which was canceled. Instead, Meta focused on Avocado, using competitor models for its training, such as Google Gemma, solutions from OpenAI, and even Qwen from China’s Alibaba, which caused surprise among insiders, as Mark Zuckerberg previously publicly criticized Chinese AI technologies due to the risk of censorship.

The company also focused on personnel changes. A key role in the development of the new strategy is played by Alexandr Wang, founder of Scale AI, who, after a $14.3 billion deal, joined Meta and closely collaborates with Zuckerberg in the TBD Lab division. Wang, known for his preference for closed models, actively develops the new course of the company, while the “godfather of AI” Yann LeCun left Meta, as his vision did not align with the new strategy.

Meta continues to focus on the development of artificial intelligence, reducing expenses on the metaverse and virtual reality, concentrating resources on AI and smart glasses. This indicates that artificial intelligence has become the number one priority for the company.