Apple’s new Mac minis are primed for native AI. I specced one for $2,899

Now hip to the Mac mini’s native LLM prowess, Apple is again with new and souped-up Mac minis that swap out the older M4 and M4 Professional chips with mightier M6 and M5 Professional chips (confusingly, the M5 Professional processors are the extra highly effective ones) providing extra CPU and GPU cores, greater reminiscence bandwidth (as much as 160GB/s), and most significantly, a strong quantity of unified RAM, which is the be-all-end-all of speedy AI inference.

However whereas the latest Mac minis are nonetheless less expensive than AMD Ryzen AI Halo and Nvidia RTX Spark programs (which, arguably, are literally competing with Apple’s stratospherically priced Mac Studio), their costs have spiked in comparison with their two-year-old M4-packing predecessors. The reason being… properly, you realize why.

Apple simply introduced its new M6 and M5 Professional Mac minis, so we haven’t had the possibility to benchmark them in comparison with the older M4 and M4 Professional fashions, nor have we stacked them up in opposition to AMD’s and Nvidia’s native AI workhorses. For its half, Apple is promising an total 4x native AI enhance versus the older Mac mini fashions, however we haven’t put that declare to the take a look at but. (There are additionally some vital “gotchas” to contemplate with native AI on a Mac mini, which I’ll get to in a bit.)

What I can do is let you know the Mac mini system I’d be eyeing as a neighborhood AI person, assuming I have been sport for getting a premium Apple system on the very top of the PC {hardware} market—which I’m not, and also you in all probability shouldn’t be both. However hey, at the least we are able to browse, proper?

For background, I at the moment personal a Mac Mini M4 Professional with 64GB of unified RAM and 512GB of inner storage, with that storage upped by a number of exterior SSDs plus a 1GB Thunderbolt NVMe for loading native AI fashions. (I ought to’ve sprung for a much bigger NVMe drive, however I used to be being low cost.)

My explicit Mac mini value me roughly $2,000 in late 2024, and whereas it could actually deal with the present native LLM favourite, Alibaba’s Qwen 3.8 27B, the MLX variant of that mannequin (optimized to run on Apple Silicon) begins to chug as soon as its context window fills to a lot greater than 40,000 tokens or so, making the Mac mini’s cooling followers spin up like jet engines.