Meta to open source its most powerful AI model as it takes swipe at OpenAI, Anthropic
Meta launched Muse Glimmer and plans to release Muse Spark 1.2 weights as Zuckerberg pushes for U.S. leadership in open AI.
Meta's announcement that it will release the weights for Muse Spark 1.2 is not an act of charity. It is a strategic bid to set the default standard for open AI development, and to position itself as the adult in the room while its rivals keep their models behind closed doors.
The timing matters. Zuckerberg frames the move around U.S. leadership in AI, a useful patriotic wrapper for what is also a commercial calculation. Open weights attract developers, and developers attract a moat of third-party tools, fine-tuned variants, and real-world deployment data that no proprietary model can easily match.
The swipe at OpenAI and Anthropic is deliberate. By casting them as closed and therefore less trustworthy, Meta invites regulators, researchers, and enterprises to see open sourcing as the responsible choice. That framing quietly converts a business decision into a policy position, and it gives Meta leverage in the coming fight over AI governance.
Releasing a model after already shipping an earlier version, Muse Glimmer, also de-risks the narrative. Meta can claim momentum, gather community feedback before the larger release, and correct course before its most powerful weights ever reach the public. The order of operations is as instructive as the announcement itself.
None of this is a referendum on whether open AI is good or bad. It is simply the latest move in a market where control over AI infrastructure is the prize, and where openness has become a competitive weapon. Meta is choosing the weapon it thinks wins.