Meta Considers Shifting from Open-Source AI Model Behemoth to Proprietary System

Meta is reevaluating its open-source strategy for its advanced AI model, Behemoth. Senior executives, including newly appointed chief AI officer Alexandr Wang, recently debated halting its public release in favor of developing a closed-source alternative. This marks a potential pivot for Meta, long lauded for its commitment to open AI development.

The company has historically argued that open-sourcing AI code accelerates innovation and democratizes access. Yann LeCun, Meta's top AI researcher, previously championed this approach, stating open platforms WOULD ultimately dominate. Chinese firm DeepSeek's successful chatbot built on Meta's public code exemplifies the tangible benefits of this philosophy.

Internal evaluations reportedly showed Behemoth underperforming against benchmarks despite extensive training. The project's pause follows last month's establishment of Meta's superintelligence division, suggesting strategic recalibration at the highest levels. No final decisions have been made regarding Behemoth's future.