Meta, the Silicon-Valley based social media tech giant has announced the release of a new open-weight model Muse Glimmer on Monday. Relatively smaller than the other leading AI models, Muse Glimmer has been designed for agentic tasks and can run on a PC or Mac with a single graphics card. This aims to tap on the demand for AI systems running on people’s devices.
Currently, open-weight models are cheaper than leading models from the frontier labs like OpenAI and Anthropic, and also come with publically accessible core components for easy customisation. This announcement comes at the heels of the social media giant looking to strengthen its position, and propel itself into the ongoing AI-race.
Open-weight AI is fast gaining traction as businesses now are growing wary of the ever increasing AI bills, cost, and growing security concerns involving frontier models. The recent Hugging Face, the AI-coding collaboration site hacked by a rogue OpenAI model, had used a Chinese open-weight model to defend against the attack as closed-source models have restrictions on use for cybersecurity.
Currently, open-weight models like Kimi K3 by Moonshot, Qwen3.8-Max by Alibaba, V4-Flash by DeepSeek are rivalling top systems by US AI labs. In a bid to rival these models, Zuckerberg has also requested the US to lower its barriers to reduce friction to allow further growth of American open source models. Zuckerberg has also pushed for AI model distillation or using powerful AI systems to train smaller models
In his statement, Zuckerberg also advocated for AI model distillation, or using a powerful AI system to train a smaller model. He said that Meta would implement a governance structure to give its independent directors the power to approve the safety criteria for releasing models.
