DATABASES Signal 422
Meta releases Muse Glimmer, a new open-weight model, and will release an open-weight version of its most advanced model, Muse Spark 1.2, in the coming weeks (Meghan Bobrowsky/Wall Street Journal)
↗
Meta has open-sourced a 30-billion-parameter model called Muse Glimmer that can run on a single consumer GPU, and promises an open-weight version of its newer Muse Spark 1.2 soon.
Engineers can now host a capable, agentic language model on-premises, eliminating reliance on external API services and the associated latency and data-privacy concerns. The open-weight release also lets teams fine-tune the model on proprietary data, integrating it directly with internal databases and workflows.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
Muse Glimmer is quantized to under 20 GB and runs on GPUs with about 24 GB of VRAM, enabling local, always-on agentic tasks.
Open weights under an Apache 2.0 license allow custom fine-tuning and embedding of the model in proprietary systems.
The upcoming open-weight Muse Spark 1.2 will extend these capabilities, but is not yet available.
THE CLUSTER