Meta’s Superintelligence Lab Considers Shifting A.I. Strategy
Meta is evaluating significant changes to its artificial intelligence approach, signaling a potential shift in its long-standing strategy. Key figures in the company’s newly formed superintelligence lab, including Alexandr Wang, Meta’s new chief A.I. officer, are discussing the possibility of moving away from their most powerful open source A.I. model, known as Behemoth, in favor of a closed model.
From Open Source to Closed Models
Historically, Meta has favored open sourcing its A.I. models, making the code publicly available for developers worldwide. This approach has been a cornerstone of their philosophy, with the belief that open development accelerates innovation and broadens access.
Switching to a closed model means keeping the underlying code private. This would mark a fundamental change for Meta, which has earned praise for its openness. One of its prominent A.I. executives previously stated that “the platform that will win will be the open one.” Open sourcing has even enabled companies like China’s DeepSeek to build advanced A.I. chatbots using Meta’s code.
Current Status of Behemoth and Lab Discussions
Meta completed the data training phase for Behemoth but has delayed releasing it due to underwhelming internal performance. Following the announcement of the superintelligence lab, teams working on Behemoth have paused further testing.
The ongoing discussions at the superintelligence lab are preliminary. No final decisions have been made, and any strategic changes would require approval from Meta’s CEO, Mark Zuckerberg. It’s possible Meta could maintain its open source models while prioritizing a closed alternative.
Strategic Implications
This potential pivot comes as Meta seeks to remain competitive against other major A.I. players such as Google, OpenAI, and Anthropic. Opting for a closed model could impact how the company collaborates with the broader developer community and how quickly its technology evolves.
Executives and strategists should watch closely as this situation develops, as it may signal a shift in how leading tech companies balance openness with proprietary development in A.I.
Additional Context
- Meta recently invested $14.3 billion in Scale AI, an A.I. startup founded and led by Alexandr Wang.
- Behemoth is classified as a “frontier” model, representing Meta’s most advanced open source A.I. effort.
For those interested in staying ahead on A.I. strategy and development, exploring targeted A.I. courses by job function can provide valuable insights into emerging trends and technology management.
Your membership also unlocks:
