Meta is fully shifting towards closed-source models, with its new model, Avocado, expected to launch next spring.

On December 10th, it was reported that after investing hundreds of billions of dollars and assembling the most expensive R&D team in tech history for several months, Meta CEO Mark Zuckerberg is personally involved in the company's day-to-day R&D work and is accelerating the overall strategic shift towards commercially viable AI models.

According to multiple sources, a new large-scale model, internally codenamed "Avocado," is expected to officially debut in the spring of 2026, and is highly likely to be released in a closed-source format—meaning Meta will exercise strict control over the model and sell access to external customers.

This strategic shift is seen as a major departure from Meta's long-standing and widely advocated open-source approach, and marks a crucial step forward in the company's commercialization of artificial intelligence.

It is understood that Zuckerberg has devoted a significant amount of time to the core team called TBD Lab, personally participating in the formulation of technology roadmaps and product strategies.

During the training of the "Avocado" model, the team innovatively integrated third-party model architectures from several competitors, including Google's Gemma, OpenAI's GPT series of open-source models, and Alibaba's Qwen, attempting to build a competitive hybrid model system by combining the advantages of diverse technologies.

Meanwhile, Meta is comprehensively reconfiguring its resource allocation, significantly reducing long-term investments in the metaverse and virtual reality fields, and instead shifting its funding and manpower towards artificial intelligence hardware, such as the ongoing AI smart glasses project.

Furthermore, the company plans to invest approximately $600 billion in the US over the next three years to build a global high-performance AI infrastructure to support its growing demand for large-scale model training and inference.

This strategic shift not only reflects Meta's pressure to seek faster commercial returns in the fierce AI race but may also have a profound impact on its long-cultivated "open ecosystem" image.

Industry observers point out that if Avocado is ultimately released as a closed-source application, it will further exacerbate the divergence between open-source and closed-source approaches in the global large-scale model market, and may reshape the landscape of corporate cooperation and competition in AI technology.