Meta released the lightweight AI model Muse Glimmer, which is available for free download as an open-source model and can run locally on a single card

According to Zhitong Finance APP, Meta (META.US) has launched a lightweight new artificial intelligence model that is small enough to run on a single computer, allowing users to download and customize this technology.

Meta announced on Monday that the new model, named Muse Glimmer, is a distilled version of the company's Muse Spark 1.2 model, designed with a focus on enhancing efficiency to minimize system configuration requirements.

According to Meta, Muse Glimmer has 30 billion parameters and is "small enough" to be driven by a single graphics card (GPU), with its main application scenarios focusing on AI tasks related to agents, such as schedule management and file organization.

The weights of Muse Glimmer (i.e., the numerical parameters that help the AI system make decisions) will be available for download on Hugging Face. Meta also plans to provide weights for a more powerful version of Muse Spark. Downloading will be free, but users will need to bear the computing costs themselves.

This initiative to open-source the AI model through a licensing agreement echoes the market competition model in China—where startups like DeepSeek and Yuezhianmian, as well as internet giants like Alibaba, have adopted this approach to attract the largest user base.

Meta CEO Mark Zuckerberg stated in a broad discussion article containing policy suggestions released on Monday, "Rather than centralizing superintelligence, we should distribute it widely, empowering everyone to control it," adding, "This is expected to usher in a new era of personal empowerment, allowing individuals to leverage this powerful new capability to fully unleash their potential."

In comparing Meta's approach with that of American competitors that "focus on building AI for businesses, governments, or other institutions," Zuckerberg emphasized that Meta is creating "personal superintelligence for everyone."

The company is currently in a global race to achieve breakthroughs in AI technology and deliver practical benefits, thereby providing returns on its massive investments in hardware and data centers. Alongside American peers like Amazon, Alphabet (Google's parent company), and Microsoft, Meta has committed to investing over $20 trillion to expand its computing power—developing AI models and tools that attract users away from competitors will help secure the company's long-term revenue. Last month, the company began charging fees for developer access to its Muse Spark 1.1 model for the first time.

Meta is also formulating a cloud infrastructure business plan to sell access to AI computing power and models in a manner similar to Amazon Web Services (AWS). However, since Zuckerberg has not disclosed more details about the plan, investors concerned about the company's return on investment (ROI) have expressed frustration.

On Monday, Meta announced plans to establish a $1 billion fund to invest in local communities in the U.S. where it owns and operates data centers. Due to opposition from critics regarding land use, energy consumption, and water resource depletion by data centers, Meta and other hyperscale computing companies like Google have faced resistance in various regions of the U.S