Apple launches first M6 Mac Mini with 4X faster AI performance for always-on AI agents

Apple is turning the Mac mini into something bigger than a tiny desktop computer. On Tuesday, the company unveiled its first M6-powered Mac mini, pitching the machine squarely at developers running AI models and autonomous agents locally rather than sending every task to the cloud.

Summary Highlights

• Apple unveiled its first M6-powered Mac mini, positioning the tiny desktop for local AI workloads and always-on AI agents.
• The M6 Mac mini delivers up to 4X faster AI performance and up to 4.8X faster LLM prompt processing in LM Studio than the M4 model.
• The new Mac mini starts at $899, with M6 and M5 Pro configurations available.
• Apple refreshed the Mac Studio with M5 Max and M5 Ultra chips, with M5 Ultra systems that can work together to run trillion-parameter AI models.
• The new Mac mini and Mac Studio are available for preorder now and begin shipping September 22.

The new Mac mini pairs Apple’s M6 chip with up to four times faster AI performance, twice the graphics and storage performance, and a 40% faster CPU compared with the previous generation, according to Apple. It starts at $899, up from $799 for the prior model, and goes on sale September 22.

The timing says plenty about where Apple thinks personal computing is headed. Developers are increasingly using dedicated Macs to run AI agents around the clock, and the Mac mini’s small footprint, low energy consumption, unified memory architecture, and Neural Engine make it an unusual fit for that job.

Apple has a name for the use case: “always-on, deskside agentic computing.”

That phrase may prove more significant than the new chip itself.

The Mac mini is becoming a personal AI server

For decades, desktop computers were machines people actively sat in front of and operated. AI agents are starting to challenge that model. A computer can now sit on a desk and continuously execute software tasks, interact with services, process information, write code, and respond to instructions with little direct supervision.

Developers building agents with software such as OpenClaw have increasingly turned to dedicated Mac minis instead of cloud infrastructure. Keeping models and data local can offer greater privacy and remove recurring cloud-compute charges. A physical machine can stay online around the clock without tying up someone’s primary laptop.

Apple appears to see an opening.

“Mac mini has always been our most versatile Mac. Whether it’s being used as a home computer, powering a professional studio, or as an always-on agentic device, it’s the little Mac that can do it all,” Apple hardware chief Johny Srouji said.

The M6 marks Apple’s first 2-nanometer Mac chip and is manufactured by Taiwan Semiconductor Manufacturing Co. Apple says its Neural Engine and unified memory architecture allow developers to run and fine-tune large AI models locally.

“With these frameworks and new chips, developers can run and fine-tune large AI models locally on their Mac,” Apple said.

Buyers can choose between M6 and M5 Pro configurations. Apple says the M5 Pro version can process large language model prompts 8.5 times faster than older Pro-equipped Mac minis.

Connectivity is also getting an upgrade, with Wi-Fi 7, Bluetooth 6, 2.5Gb Ethernet, and an optional 10Gb Ethernet connection. Those additions could matter more for machines acting as persistent AI nodes than they would for an ordinary home desktop.

Apple pushes Mac Studio deeper into local AI

Apple is applying the same strategy higher up the performance ladder.

The refreshed Mac Studio comes with M5 Max or M5 Ultra chips and targets developers and researchers working with much larger models. Apple says the M5 Max can run LLMs nearly four times faster than its predecessor.

The more interesting configuration is M5 Ultra. Multiple Mac Studios equipped with the chip can connect and pool their memory, giving researchers enough local compute to run trillion-parameter frontier AI models, according to Apple.

Apple says the M6 Mac mini can process LLM prompts in LM Studio up to 13.5 times faster than the M1 Mac mini and 4.8 times faster than the M4 model. The company says it can also deliver up to twice the gaming performance with ray tracing compared with the M4.

That blurs the line between a desktop workstation and AI infrastructure.

Mac Studio with M5 Max starts at $2,499, unchanged from the previous generation. The M5 Ultra model starts at $5,499, up from $5,299 for the previous Ultra configuration.

Photo credit: Apple

Apple’s Mac strategy comes at an interesting moment. Much of the AI industry’s attention has centered on enormous GPU clusters, hyperscale data centers and cloud inference. Yet another computing model is forming at the opposite end of the spectrum: AI running locally on machines owned by individuals and businesses.

That model plays directly into Apple’s strengths. The company controls its chips, hardware, operating system and AI frameworks, giving it unusually tight control over how models interact with the machine.

The Mac mini could be one of the clearest examples of that strategy.

Apple once sold the tiny computer as an inexpensive way to get a Mac without buying a display. In the agent era, the more interesting question may be whether people need to sit in front of it at all.

An $899 Mac mini running AI agents day and night could become less of a PC and more of a personal AI server.

And Apple appears increasingly comfortable selling it that way.