Top members of Meta’s new Superintelligence Lab are reportedly discussing a shift away from the company’s powerful open-source AI model, Behemoth, in favor of developing a closed model, according to The New York Times.
Sources indicated that while Meta has completed training on Behemoth, the model’s release was delayed due to disappointing internal performance. Following the launch of the Superintelligence Lab, testing on Behemoth reportedly came to a halt.
These discussions are preliminary and would require approval from Meta CEO Mark Zuckerberg. A company spokesperson told TechCrunch that Meta’s stance on open-source AI remains “unchanged.” They stated, “We plan to continue releasing leading open-source models. We haven’t released everything we’ve developed historically and expect to continue training a mix of open and closed models going forward.”
The spokesperson did not comment on the potential shift away from Behemoth. A move towards prioritizing closed-source models would represent a significant philosophical change for the company.
While Meta utilizes advanced closed-source models internally, such as those powering its Meta AI assistant, Zuckerberg has positioned open-source as a core part of its external AI strategy. He has publicly criticized competitors like OpenAI for becoming more closed after partnering with Microsoft, emphasizing the openness of the Llama family as a differentiator.
However, Meta faces pressure to monetize beyond advertising as it invests billions in AI, including offering lucrative signing bonuses and high salaries to attract top researchers, expanding data centers, and covering the substantial costs associated with developing artificial general intelligence (AGI) or “superintelligence.”
Despite housing one of the leading AI research labs globally, Meta still lags behind rivals like OpenAI, Anthropic, Google DeepMind, and xAI in terms of commercializing its AI initiatives.
ICYMI: Lauren Baaba Annan Launches Landmark “Secretaries in Training” Initiative to Empower Student Leaders
If Meta prioritizes closed models, it may suggest that its commitment to openness was more strategic than ideological. Past comments from Zuckerberg reveal some ambivalence about fully committing to open sourcing Meta’s models. He mentioned on a podcast last summer, “We’re obviously very pro open source, but I haven’t committed to releasing every single thing that we do.”
Closed models would provide Meta with greater control and monetization opportunities, particularly if it believes its acquired talent can deliver top-tier performance.
Such a shift could reshape the AI landscape, potentially slowing the momentum of open-source initiatives largely driven by Meta and models like Llama. Power could return to major players with closed ecosystems, while grassroots open-source development might continue to thrive among smaller companies focused on fine-tuning, safety, and model alignment.
On a global scale, Meta’s retreat from open source could allow China, which has embraced open-source AI initiatives like DeepSeek and Moonshot AI, to strengthen its domestic capabilities and global influence.
SOURCE: TECH CRUNCH
