Apple's Mac Studio Now Runs Powerful AI Locally, No Cloud Needed

Apple has released a new version of its MLX software that lets Mac Studio computers run huge AI models right on your desk. These models, like the nearly trillion-parameter Muse Spark Open, are normally too big for personal computers. But with MLX and Apple's powerful M5 Ultra chip, a single Mac Studio can handle models up to 128 billion parameters, and a cluster of four can run the giant Muse Spark Open at over 36 tokens per second—faster than average human reading speed. This means you can have a private, local AI assistant without relying on the internet or cloud services.

The benefit for you is privacy and cost. Running AI locally means your data never leaves your device, so sensitive information stays secure. You also don't need to pay for cloud computing, which can add up. Plus, the Mac Studio is energy-efficient, using about as much power as a toaster, so it won't spike your electricity bill. This is great for developers, researchers, or anyone who wants to experiment with AI without worrying about data leaks or ongoing fees.

The catch? The largest models require a lot of memory—Muse Spark Open needs over 1.6 terabytes of RAM, which is more than a single Mac Studio offers. So you'd need to link multiple machines together, which is more complex and expensive. Also, while the speed is impressive, it's still slower than top-tier cloud AI for some tasks. But for many practical uses, like coding assistance or private chatbots, it's more than enough.

Overall, this is a big step toward making powerful AI accessible and private for individuals and small businesses. It shows that you don't always need a massive data center to run cutting-edge AI—sometimes, a high-end desktop computer can do the job.