Meta releases Muse Glimmer, 30B-parameter open-weights model for PCs

Meta's release of Muse Glimmer follows a broader industry trend toward democratizing large language models by making them available in open-weights form—meaning the model's internal weights are publicly accessible, allowing researchers and developers to inspect, modify, and run them independently. The 30 billion parameter size positions the model in the mid-range of modern language models, large enough for meaningful tasks but small enough to run on consumer-grade hardware without specialized accelerators. This contrasts with proprietary API-only models, which require cloud resources and vendor intermediation. Zuckerberg's accompanying essay on AI risks and open-source regulation suggests Meta is positioning open-weights release as a policy argument—that transparency and accessibility, rather than centralized control, better serve long-term AI safety and innovation.