
Meta on Monday released Muse Glimmer, a compact open-weight AI model designed for agentic tasks that can run on a Mac or PC with a single graphics card. The 30-billion-parameter model is…
Meta on Monday released Muse Glimmer, a compact open-weight AI model designed for agentic tasks that can run on a Mac or PC with a single graphics card. The 30-billion-parameter model is compressed to about 4-bit precision, bringing it under 20GB, and supports over 100 languages. It can handle coding, function calling, local file management and other multi-step workflows, and the weights are available on Hugging Face.

CEO Mark Zuckerberg used the launch to renew his call for fewer US restrictions on open-source AI, arguing that foreign labs, especially Chinese ones, hold advantages because American labs face additional training-data restrictions. He also defended AI distillation. The launch comes as Chinese startups like Alibaba, DeepSeek and Moonshot gain ground with open-weight models. Meta's shares rose 1 per cent in premarket trading, Reuters reported.

Zuckerberg's pitch for open-source AI is not pure philanthropy. Meta benefits when open models lower costs for everyone, and his warning about Chinese rivals serves his business interest. But his point about US restrictions on training data is real, and closed models from OpenAI and Anthropic are not inherently safer. The test will come with Watermelon, Meta's next frontier model. If it stays closed, Zuckerberg's openness talk is just strategy, not philosophy.
Sources (4): ndtvprofit.com, hindustantimes.com, timesnownews.com, livemint.com
This story was synthesised by AI from the 4 sources linked above.
Updated: this story now draws on 4 sources.