Apple refreshed the Mac mini and Mac Studio with the M6 and M5 Ultra for local model work. The M6 is the first 2-nanometre Mac chip, while the M5 Ultra offers 36 CPU cores, 80 GPU cores and up to 512 GB of unified memory. macOS 26.2 enables MLX distributed inference across Thunderbolt 5 hosts. The machines ship on 22 September.
Apple refreshed the Mac mini and Mac Studio with two new chips aimed at local model work. The M6 is the Mac line's first 2-nanometre chip. The M5 Ultra offers 36 CPU cores, 80 GPU cores and up to 512 GB of unified memory.[1]
macOS 26.2 had already enabled distributed inference with MLX across Thunderbolt 5 hosts. Developers can therefore run open-weight models too large for one machine across more than one Mac.[1]
The new machines are available for preorder and ship on 22 September. The announcement presents that timetable as the hardware's market-release plan, not as a benchmark of local-inference performance. The announced core and memory specifications of the M6 and M5 Ultra show the focus on local model work on one machine. For larger open-weight models, the route identified in the report is to use more than one Mac with MLX over Thunderbolt 5. Apple refreshed the Mac mini and Mac Studio with the M6 and M5 Ultra for local model work. Earlier MLX distributed-inference support preserves a multi-Mac option for open-weight models too large for one machine. This briefing keeps together the actor, reported scope and stated uncertainties of the event described by the source. It adds no separate development, definite outcome or comparison absent from that source; the headline isolates only this event.[1]
