Meta announced its open-weight model “Muse Spark Open” on Thursday, local time. It has 816 billion parameters and a context window of 1 million tokens, and is designed for local use on Apple computers. It is already available through several API platforms.
It runs on a Mac—but not necessarily your Mac
When people hear “AI that runs on a Mac,” they may picture downloading it to the laptop they use every day and giving it a try. But this is not so straightforward. Reports say running a model of this size—816 billion parameters—locally requires at least 1TB of memory. That’s an order of magnitude beyond the memory in a typical laptop.
The recommended setup is a Mac Studio with an M5 Ultra chip. Rather than making AI easier to use, what Meta has demonstrated with this model is the option to run a massive model on a device sitting on your desk, instead of entrusting it to the cloud.
Even so, the fact that this option exists is significant. Until now, ultra-large models have lived on the other side of a data center. Now they are beginning to appear within reach of a desktop computer. The hardware requirements are steep, but the direction of travel is worth watching.
From AI that writes code to AI that works with video
Muse Spark Open has a native multimodal design that can handle images and other content, and is built for complex coding tasks and agentic work. Agentic use means the AI does more than answer a question once: it plans and carries out a series of steps on its own.
Meta says it paired the model with its AI coding terminal, “Muse Code,” and used it in various agentic workflows. One example was making changes to a client for the decentralized social network Mastodon. The model handled the process from fixing a bug and adding a feature through to building the software, Meta says.
The model was also used to operate the video-editing software DaVinci Resolve, where it could review footage and crop it. From coding to professional video-editing tools, the situations in which AI operates apps on a computer screen are steadily expanding.
What I find interesting is that both examples involve work on a desktop computer. AI is moving from answering questions in a chat window to acting like a colleague who gets things done at a PC. That may be part of the rationale for focusing on local execution on Macs.
The value of open weights goes beyond performance
Based on the evaluation results it has published, Meta claims the model outperforms other leading open-weight models. It describes Muse Spark Open as the most capable open-weight model in the United States.
But the significance of open weights cannot be captured by rankings alone. Because the weights—the contents of the trained model itself—are made public, open-weight models can offer cost advantages over closed models, while also making it easier to ensure transparency about how they work and flexibility in how they are used.
For businesses, the choice is whether to use the model conveniently through an API or run it in their own environment. The former is quick to deploy; the latter avoids sending data outside the company. The 1TB requirement is a high bar, but organizations handling sensitive development work or video footage may decide the investment is worthwhile.
Muse Spark Open gives organizations one more factor to consider. Try it through an API, then bring it in-house if needed. This two-step approach may become the standard way to work with open-weight AI.
Comments