Meta has released Muse Glimmer, an artificial intelligence model with 30 billion parameters. This model can run on a single consumer graphics card.
The company made it available as a free download on Hugging Face under the Apache 2.0 license. This allows developers to use and modify the model for commercial purposes with few restrictions.
Muse Glimmer is based on Meta’s larger Muse Spark 1.2 model but has been compressed to focus on tasks such as organizing files, coding on local machines, and managing schedules. Developers reduced the model’s size to under 20 GB using 4-bit precision, so it can run on typical graphics cards with 24 GB or 32 GB of memory.
To improve speed, the system uses a special network called DFlash. This helps the AI predict several words at once rather than one by one. Tests within the company showed that this method is much faster on NVIDIA and Apple chips than older methods of generating text.
Along with the software release, Meta CEO Mark Zuckerberg wrote a 14-page essay supporting open-source technology. He said it is safer for the public to have access to advanced AI than to have only a few large companies control it.
Zuckerberg also mentioned that strict US policies may make it harder for American developers to compete with Chinese AI companies such as Alibaba and DeepSeek.
READ MORE: WhatsApp adds new feature to hide names while voting in polls
Security remains a central issue in the debate over open and closed technology. Recent events, such as a vulnerability in OpenAI’s systems and an accidental internet exposure of Meta’s Muse Spark 1.1 during testing, have raised concerns about the unpredictable risks associated with autonomous software.
Despite these security challenges, Meta is continuing to push forward with its open-source strategy. The company said it plans to eventually release the main weights for Muse Spark 1.2 to the public, moving away from its recent focus on paid, proprietary models.
After the announcement, Meta’s stock rose nearly 3 percent in early trading, reflecting investor confidence in a more open approach to artificial intelligence.