Mac Mini (M6) review: still great, but a tougher sell
There are two chip options for the 2026 Mac mini: the M6 and the M5 Pro. The M5 Pro is the more confusing of the pair, as Apple sells it in two configurations. The $1,699 base model has a 15-core CPU – five of Apple’s fastest “super” cores plus ten performance cores – with a 16-core GPU featuring integrated neural accelerators in each core, hardware-accelerated ray tracing, and a 16-core neural engine. Step up and you get an 18-core CPU and a 20-core GPU. Memory bandwidth is 307 GB/s in both cases, almost double the M6’s best figure, and that’s what really separates the two machines for anyone running local AI models, which has become a major use case for the Mac mini. The memory options also go further. The M5 Pro starts with 24GB of unified memory and configures to 48GB or 64GB, double the M6’s cap.
Many, however, will opt for the M6 variant instead. The M6 contains a 12-core CPU, a 12-core GPU with neural accelerators on each GPU core, and a dual 16-core neural engine. Base memory is 16GB of unified memory – expandable to 24GB or 32GB – with up to 170GB/s bandwidth on higher memory configurations and 153GB/s on the 16GB base. Storage starts at 256GB and tops out at 2TB.
I just finished reviewing the M5 Ultra Mac Studio and put it through its paces for local AI use. The M6 Mac mini is a decidedly different machine. Yes, it can be useful for local AI, but its use for that purpose is much more limited than the workhorse that is Mac Studio. On the Mac mini M6, you can run small local templates to perform specific, defined tasks, but you won’t necessarily use general-purpose templates to replace services like ChatGPT.
For use without AI, the M6 is very fast. App launches feel instant, multitasking is effortless, and workflows involving coding, photo editing, and light video work are pretty snappy. Apple claims up to 40% faster CPU performance, 2x faster graphics, and 4x faster AI, plus workload-specific gains like a 4.8x faster LLM prompt in LM Studio and 1.5x faster Excel calculations.
Of course, this is the base model, and its performance still has limitations compared to high-end machines. Demanding video editors, or even those who need more than the basics, will probably want to upgrade to at least an M5 Pro. Plus, as noted, local AI enthusiasts will need to find something with more RAM.
The new Mac mini also performed well in performance tests. In Geekbench 7, it performed about 25% better than the M4 model in single-core scores and even better in multi-core tests, and actually performed similar to the M5 Pro machines in single-core performance – although the M5 Pro was well ahead in multi-core. More importantly, it’s more powerful than anything on the Windows side when it comes to compact PCs, especially in terms of CPU performance.
But again, it’s a much tougher sell at this price point than the $599 M4. It’s certainly a great machine, but the value isn’t as impressive. Fortunately for Apple, the Mac mini was so far ahead in terms of value that the M6 model is, frankly, still well worth its premium price.
