Meta Platforms introduces new open source AI models, signalling a commitment to wider access amid industry tension over control and commercial interests, as Mark Zuckerberg advocates for broader availability of advanced AI technology.
Meta Platforms has doubled down on open source in artificial intelligence, unveiling a new model that it says developers can run on a personal computer while Mark Zuckerberg warned that the technology’s benefits will be reduced if advanced systems are controlled by a narrow group of companies, governments or institutions.
According to the announcement, the company has released Muse Glimmer under a permissive licence, a move that allows outside software makers to use and adapt the model more freely than they could with a tightly restricted product. Zuckerberg also said Meta will offer access to a more powerful system, Muse Spark 1.2, extending the company’s effort to keep at least part of its AI stack available to developers rather than locked inside its own services.
The release comes after a difficult period for Meta in the AI race. Reuters and other outlets have reported that the company has struggled to match rivals such as OpenAI, Anthropic and Google, even as it has positioned open-source development as a strategic advantage. TechCrunch reported earlier this year that Muse Spark emerged from a broader overhaul of Meta’s AI work and was being integrated into the company’s app and websites, with later roll-outs planned across Facebook, Instagram and WhatsApp. Other reports have described Muse Spark as a more proprietary turn, highlighting the tension between Meta’s public support for openness and the commercial reality of competing with closed models.
In a lengthy essay published online, Zuckerberg argued that broad access to advanced AI would be better for individuals than a system dominated by large organisations. He said the United States and its allies must move faster on energy capacity and infrastructure to keep pace with China, and he defended “distillation”, the practice of teaching a smaller model from the output of a stronger one. That position is notable because the Trump administration has signalled tougher action against Chinese AI firms accused of using U.S. models in that way, while Anthropic has previously made a similar accusation against DeepSeek.
The broader policy message was clear: Meta wants AI to remain widely usable, not concentrated in a few hands. Whether that vision can coexist with the company’s own competitive ambitions remains the central question, especially as the industry continues to shift towards larger, more capable systems and tighter control over distribution.
Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.
