Meta's AI team weighs scrapping open‑source model for closed alternative
What's the story
Meta's newly established Superintelligence Lab is contemplating a major shift in its artificial intelligence (AI) strategy. A team including Alexandr Wang, Meta's new chief AI officer, is considering replacing the company's most powerful open-source AI model, Behemoth, with a closed one. This would be a major departure from Meta's traditional practice of open-sourcing its AI models and making their code publicly available for other developers to use.
Model performance
Meta had already trained its Behemoth model
Meta had already trained its Behemoth model by feeding it data to improve its performance. However, the company decided to delay the model's release due to underwhelminh performance in internal tests. After announcing the formation of the Superintelligence Lab last month, teams working on Behemoth, a "frontier" model, reportedly stopped testing it altogether.
Strategic shift
Discussions still in early stages
The potential shift toward a closed AI model would be a major philosophical change for Meta. The discussions within the Superintelligence Lab are still in their early stages and no decisions have been made yet. Any such changes would require approval from Meta CEO Mark Zuckerberg, who has so far maintained that the company's stance on open-source AI remains "unchanged."
Future plans
'We expect to continue training mix of open, closed models'
A Meta spokesperson recently reiterated the company's commitment to releasing leading open-source models. The spokesperson said, "We haven't released everything we've developed historically and we expect to continue training a mix of open and closed models going forward." However, they did not comment on the potential shift away from Behemoth. If this happens, it would mark a major departure for Meta as it looks to prioritize closed-source models.
Market pressure
Meta is under pressure to monetize beyond ads
Meta is under pressure to monetize beyond ads as it invests billions in AI. This includes paying huge signing bonuses and nine-figure salaries to attract top researchers, building new data centers, and covering the high costs of developing artificial general intelligence (AGI), or "superintelligence." Despite having one of the world's top AI research labs, Meta still lags behind competitors such as OpenAI, Anthropic, Google DeepMind, and xAI in commercializing its AI work.
Strategic implications
If Meta prioritizes closed models, it could impact AI landscape
If Meta prioritizes closed models, it could indicate that its commitment to openness was more strategic than ideological. Closed models would give the company more control and monetization opportunities, especially if it believes its acquired talent can deliver competitive performance. Such a shift could also impact the AI landscape, potentially slowing open-source momentum and shifting power back toward major players with closed ecosystems.