Hajra Naz
Meta has launched what it calls the “most capable” open-weight model in the United States. The model is named Muse Spark Open. It was also built to work with Macs.
Key Specs and Claims
The new model uses 816 billion parameters. It has a 1 million token context window. Meta says it outperforms other popular open-weight models. The claim rests on evaluation results Meta released on Thursday. Those results come from Meta’s own testing.
What It’s Built For
Muse Spark Open is natively multimodal. It’s built for complex coding tasks and agentic work. It’s also designed to run locally on Apple computers.
The Hardware Requirement
Running it locally takes serious power. According to Engadget, you need at least 1 terabyte of RAM. The model runs best on a Mac Studio with the M5 Ultra chip.
Read More: Meta Launches Its First-Ever Frontier AI Model: Muse Wonder
What It Can Do
If you have the memory, the model could be a strong workhorse. It suits coding and other desktop tasks. Meta paired it with Muse Code, its AI coding terminal. The company then used it in several agentic workflows. One example was coding fixes for a Mastodon client. The work included “bug fixes, new features, [and] full build.”
Meta also used the model with DaVinci Resolve. That’s high-end video editing software. Meta says the model reviewed and cropped footage.
Why Open-Weight Models Matter
Open-weight models offer clear advantages over closed ones. They can be much cheaper to run. They also offer transparency and flexibility. Closed models don’t allow that.
Availability
Meta is making the model available now. It’s accessible through several API platforms.
