Meta Platforms Inc. released an open-source AI model called Muse Glimmer on Aug. 10 that can run locally on a laptop [1, 2].
The release marks a shift toward decentralized artificial intelligence. By removing the requirement for cloud connectivity, the model allows users to process data privately and avoid the subscription fees associated with proprietary cloud-based AI services.
Muse Glimmer features 30 billion parameters [2]. It is designed to operate on standard laptops provided they are equipped with a capable GPU [4, 5]. This hardware requirement ensures the model can handle the computational load necessary for a model of its size without relying on external servers [4].
Meta said the model is intended to democratize AI. The company said that letting users run powerful models locally reduces reliance on cloud services and lowers costs [1, 2].
The model is available for download globally under an Apache 2.0 open-source license [2]. This license allows developers and individuals to modify and distribute the software, potentially accelerating the creation of specialized local AI tools.
While some reports mentioned a different model called Muse Spark, verified data confirms the current release is Muse Glimmer [2]. The model is now available for public deployment [1].
“Muse Glimmer features 30 billion parameters.”
The shift toward high-parameter models running on consumer hardware signals a move away from the total dominance of centralized AI providers. By providing a 30-billion-parameter model under an open-source license, Meta is positioning itself as the primary infrastructure provider for local AI development, potentially forcing competitors to either open their weights or lower cloud pricing to remain attractive to developers.
