Apple is refreshing its Mac Mini and Mac Studio lineups with new 2-nanometre chips, claiming better speeds for running AI workloads locally, without sending data to the cloud.

Apple is making a case for on-device performance that ensures AI models run locally so that sensitive data does not leave the room and users are not paying per token to a cloud provider. The Mac Mini, in particular, is positioned as always-on “desk-side” compute, sitting beside a MacBook and handling automated or agentic tasks without needing to be actively used.

Highlighting memory capacity and memory bandwidth as the key difference for on-device LLMs, Apple told that the M6 Mac Mini with up to 32GB unified memory would be enough to run a “nice” local LLM, while the M5 Pro Mac Mini with up to 64GB unified memory and more bandwidth could sustain larger local models. Both new Mac Mini variants offer up to 4× AI performance over their respective prior-generation models.

Story continues below this ad

It remains to be seen whether the proposition is exciting enough for enterprises. Pre-orders open today (August 25), with availability from September 22.

Mac Mini gets a two-generation jump

The new Mac Mini ships with the M6 chip, skipping a generation over the current M4 model. The M6 has a 12-core CPU, 12-core GPU, and a dual 16-core Neural Engine. Apple claims this doubles the peak AI compute over the M4. Memory tops out at 32GB with 170GB/s bandwidth. Apple claims the M6 has 40 per cent faster multi-threaded CPU, twice the graphics speed, and four times the AI compute over the M4.

For heavier workloads, there’s an M5 Pro option that supports up to 64GB unified memory and adds Thunderbolt 5. Both variants include Wi-Fi 7, Bluetooth 6, and 2.5Gb Ethernet standard. For anyone upgrading from an M1 Mac mini, Apple claims up to 20x faster AI inference.

Mac Mini with M6 starts at Rs 99,900 and Rs 88,900 for education. Mac Mini with M5 Pro starts at Rs 2,09,900 and Rs 1,97,900 for education.

Story continues below this ad

Mac Studio goes bigger

The Mac Studio gets either an M5 Max or the new M5 Ultra. The top of the line Ultra has a quad-die chip with a 36-core CPU, 80-core GPU, up to 512GB of unified memory, and 1.2TB/s memory bandwidth. The extra memory bandwidth is what can let the machine handle very large AI models without breaking a sweat. Against the M3 Ultra, Apple claims 30 per cent faster CPU, 80 per cent faster graphics, and 4.5x higher AI compute.

The extra memory bandwidth is what can let the machine handle very large AI models without breaking a sweat.

Four Mac Studios can also be linked via Thunderbolt 5 into a cluster, which Apple says delivers up to 3x the performance of a single unit.

Mac Studio with M5 Max starts at Rs 2,79,900 and Rs 2,56,900 for education. Mac Studio with M5 Ultra starts at Rs 6,29,900 and Rs 5,84,900 for education. The 512GB memory configuration comes later, in late October.