Meta Platforms Inc. released a new artificial intelligence model called Muse Glimmer that users can run locally on consumer-grade PCs and Macs [1].
The release represents a shift toward decentralizing AI capabilities. By allowing the model to operate on personal hardware rather than exclusively in the cloud, Meta aims to reduce the reliance on a few dominant infrastructure providers.
Mark Zuckerberg said the model is part of an effort to democratize AI. He said this approach is intended to prevent the concentration of advanced AI power in the hands of a small number of companies [3, 5].
Muse Glimmer features 30 billion parameters [1]. Despite its size, the model is designed to run on a single consumer-grade GPU found in a standard laptop or desktop PC [1]. This allows users to process data and generate outputs without sending information to external servers.
Reports on the timing of the launch vary. Some sources state the model was released on Aug. 10, 2024 [1], while other reports indicate Zuckerberg announced the plans on Monday, Aug. 12, 2024 [2].
There is conflicting information regarding the accessibility of the software. Some reports state the model is open for other software developers to use [3]. However, other sources say the tools are offered through Meta platforms and are not fully open-source [6].
Meta is hosting the downloadable model globally [3]. The company said the release is a way to put powerful generative tools directly into the hands of individual users, and developers.
“Muse Glimmer features 30 billion parameters.”
The shift toward local execution of large language models reduces latency and increases privacy for the end user. By enabling a 30-billion-parameter model to run on consumer hardware, Meta is challenging the cloud-centric business models of competitors and lowering the barrier to entry for developers who cannot afford massive server clusters.
