Meta's newest open-weight AI model 'Muse Glimmer' launched - Everything you need to know

Meta, led by Mark Zuckerberg, has unveiled a new open-weight AI model called Muse Glimmer. This initiative aims to strengthen Meta's position in the highly competitive AI sector. The release of Muse Glimmer follows the establishment of a costly superintelligence team last year. In response to market optimism, Meta's shares surged by 1 percent in premarket trading, according to a Reuters report.

Muse Glimmer boasts a compact design that is considerably smaller than leading models from rival companies. It is also designed with an agentic focus, enabling it to execute complex tasks. The model runs efficiently on a Mac or PC using a single graphics card, making it accessible to a wider range of users.

One of the key advantages of open-weight models like Muse Glimmer is their lower operational costs compared to frontier labs such as OpenAI and Anthropic. These models typically offer publicly accessible core components, facilitating easier customization. The open-source nature of Muse Glimmer also allows for enhanced cybersecurity work, as demonstrated when Hugging Face turned to a Chinese open-weight model during a recent hacking incident.

Meta CEO Zuckerberg has been advocating for lower US barriers for open-source AI models to compete with Chinese rivals. His comments further bolster support for open-weight AI models. The increasing interest in these models can be attributed to businesses' concerns about the costs of AI and cybersecurity incidents. Foreign labs currently hold several advantages in this domain, as American labs face numerous restrictions on training data.

Zuckerberg emphasized that US policy must reduce this additional friction for American open-source models to lead in the long run.

Currently, Chinese startups lead the development of open-weight models, with Alibaba's Qwen3.8-Max, DeepSeek's V4-Flash, and Moonshot's Kimi K3 performing at levels comparable to some leading AI systems developed by US companies. In contrast, models from US developers such as OpenAI, Anthropic, and Alphabet's Google remain closed-source. Zuckerberg also pushed for AI model distillation, which involves using a powerful AI system to train a smaller model.