Meta released Muse Glimmer on Monday, a 30-billion-parameter open-weight AI model built for local agent work. It runs on a Mac or PC with a single consumer graphics card. The model is available under an Apache 2.0 license on Hugging Face so developers can download and run it themselves. Meta says it is optimized for tasks like function calling, local coding, and multi-step agent workflows without needing the cloud. The company plans to open the weights for a version of its larger Muse Spark 1.2 model in the coming weeks. CEO Mark Zuckerberg used the launch to call for fewer US restrictions on open-source AI. He argued that American labs face extra rules on training data that give Chinese competitors an edge. Chinese open-weight models from firms like DeepSeek and Alibaba have gained ground with lower costs. Zuckerberg said the US should aim to make its open models the best in the world.
Meta released Muse Glimmer on Monday, a 30-billion-parameter open-weight AI model built for local agent work. It runs on a Mac or PC with a single consumer graphics card. The model is available under an Apache 2.0 license on Hugging Face so developers can download and run it themselves. Meta says it is optimized for tasks like function calling, local coding, and multi-step agent workflows without needing the cloud. The company plans to open the weights for a version of its larger Muse Spark 1.2 model in the coming weeks. CEO Mark Zuckerberg used the launch to call for fewer US restrictions on open-source AI. He argued that American labs face extra rules on training data that give Chinese competitors an edge. Chinese open-weight models from firms like DeepSeek and Alibaba have gained ground with lower costs. Zuckerberg said the US should aim to make its open models the best in the world.