Meta releases Muse Glimmer, a 30B Apache 2.0 model that runs on a single consumer GPU
Meta released Muse Glimmer, a 30-billion-parameter multimodal model under Apache 2.0 that runs on one consumer GPU or a Mac with no network call. Distilled from Muse Spark 1.2 for agentic tool use and coding, it handles a 131K context and over 100 languages, and ships on Hugging Face in BF16, GGUF k-quant, and ExecuTorch. The license is the signal — Apache 2.0 at this size puts tool-using agents on local hardware with no API bill and no data leaving the machine, and Zuckerberg says the Muse Spark 1.2 weights follow in the coming weeks.
Source: engadget.com ↗
We designed Muse Glimmer to balance capability against the memory and compute constraints of local hardware.
Meta
Why this matters
- → 30B multimodal model runs on single consumer GPU with no network calls—agent tasks stay local.
- → Apache 2.0 license eliminates API bills and vendor lock-in for tool-using AI.
- → Signal that open, distributed AI is gaining parity with closed commercial models.
Superintelligence goes local