Meta’s ambitious plans for open source artificial intelligence may be entering a turning point. According to a report by The New York Times, internal discussions within Meta’s newly launched Superintelligence Lab suggest a possible shift away from its powerful open-source model known as Behemoth—a development that could signal a major change in Meta’s AI strategy.
What Is the Behemoth Model?
Behemoth is one of Meta’s most advanced AI models to date, reportedly designed to push the boundaries of large language models (LLMs). While Meta successfully completed training on the model, internal performance testing fell short of expectations. As a result, the company has delayed the release of Behemoth and halted further development since the launch of its Superintelligence Lab.
This Lab is Meta’s new initiative to compete directly with leaders like OpenAI, Google DeepMind, Anthropic, and xAI in the race to build Artificial General Intelligence (AGI).
The Closed vs. Open Source Debate
Traditionally, Meta—under CEO Mark Zuckerberg—has positioned itself as a champion of open source AI. The company’s open release of its Llama models helped power hundreds of AI startups and research projects. Zuckerberg has also been publicly critical of OpenAI for becoming more closed after its partnership with Microsoft.
However, Meta’s internal stance appears to be more nuanced. A spokesperson told TechCrunch that:
“We plan to continue releasing leading open source models… We expect to continue training a mix of open and closed models going forward.”
This indicates that Meta may not release every model it builds—especially if the capabilities raise concerns over safety or competitive advantage.
Commercial Pressure Mounts
Meta’s open source approach has fueled innovation but comes at a steep cost. The company is spending billions of dollars annually on AI infrastructure, including:
- Hiring elite AI researchers with 9-figure salaries and signing bonuses
- Constructing massive data centers
- Funding the Superintelligence Lab
Despite having one of the world’s top AI research teams, Meta is still behind in commercializing its AI tools compared to rivals like OpenAI, Anthropic, and Google DeepMind.
A shift toward closed-source models would give Meta:
- Greater control over monetization
- Enhanced data security
- Competitive edge in enterprise applications
This could also attract premium partnerships and position Meta as a provider of enterprise-grade AI services.
Global Implications of the Shift
If Meta pulls back on open sourcing its models, the effects could ripple across the tech landscape:
- Startups and Developers – Many smaller players rely on open foundation models like Llama for fine-tuning, safety research, and alignment.
- Open Source Momentum – Meta’s leadership helped open source AI flourish. A withdrawal could stall progress and centralize power.
- China’s Response – Chinese firms like DeepSeek and Moonshot AI are embracing open source as a national strategy. Meta stepping back could give China more room to expand its influence.
Even as OpenAI prepares to launch its delayed open model, Meta’s shift may realign the industry around closed ecosystems, favoring big players and making it harder for new entrants to compete.
Zuckerberg’s Ambivalence
Mark Zuckerberg’s past remarks suggest this shift is not unexpected. On a podcast in mid-2024, he stated:
“We’re obviously very pro open source… but if there’s a qualitative change in what a model can do, and we feel it’s not responsible to open source it, then we won’t.”
This pragmatic view may now shape Meta’s strategy, especially as it pursues superintelligence and AGI.
Conclusion
Meta’s decision on whether to release or shelve Behemoth will reveal much about the future of open AI. If Meta leans into closed models, it may mark the end of its role as a standard-bearer for open AI—and a signal that openness was a tactic, not a commitment.
As costs rise, risks increase, and global pressure mounts, companies like Meta are navigating a delicate balance between innovation, safety, and profit. One thing is clear: the next move from Meta will shape not just its own future, but the future of open AI worldwide.
Related Articles: