Apple launches Mac Studio with M5 Ultra chip targeting on-device AI workloads

The base Mac mini, by contrast, ships with an M6 chip manufactured on a 2-nanometer single-die process. The M6 integrates 12 CPU cores (2 super, 4 performance, 6 efficiency), 12 GPU cores with Neural Accelerators, and a dual 16-core Neural Engine. Memory tops out at 32 GB with 170 GB/s of bandwidth, and Apple positions the M6 Mac mini as suitable for medium-sized models of roughly 30 billion parameters. The M6's GPU AI throughput is rated about 30 percent above the prior M5 generation.

Pricing across the lineup

Pricing separates the two product lines sharply. The M6 Mac mini starts at 899 dollars in the United States and 1,049 euros in Germany, Austria and Switzerland. The M5 Ultra Mac Studio starts at 5,499 dollars in the US and 6,599 euros in DACH markets. Apple also offers a mid-tier M5 Pro Mac mini from 1,699 dollars / 1,999 euros and an M5 Max Mac Studio from 2,499 dollars / 2,999 euros, giving buyers four entry points across the refreshed lineup.

Apple is pitching the M5 Ultra Mac Studio primarily at developers and researchers who want to run large language models locally rather than relying on cloud infrastructure. The combination of 512 GB of unified memory and 1.2 TB/s of bandwidth is intended to keep multi-hundred-billion-parameter models resident in memory, avoiding the latency and cost of round trips to remote servers. By comparison, Apple's prior M3 Ultra generation is rated at roughly one-fifth the AI compute throughput of the new chip, according to the figures Apple published.

Targeting AI workloads on-device

The Mac mini refresh introduces the M6 family at a more accessible price point. With 32 GB of unified memory, the base model can host AI models in the tens-of-billions-of-parameters range, a category that includes many open-source assistants and image generators. Apple says the M6's GPU offers Neural Accelerator support across all 12 cores, a step up from earlier chips where acceleration was limited to a subset of cores.

Apple did not detail in the announcement which software frameworks beyond its own MLX stack would target the new chips, though the company has historically supported TensorFlow and PyTorch through its ML Compute and Metal backends. The Genlock synchronization standard cited in technical documentation defines a method for line- and frame-accurate synchronization of multiple video signal sources, a feature relevant to professional video production workflows running on the Mac Studio.

Availability and supply

Industry analysts note that the pricing gap between the Mac mini and Mac Studio lines has widened with this generation. The flagship M5 Ultra configuration costs roughly 6.3 times the entry-level M6 Mac mini in DACH markets, compared with a smaller multiplier in earlier Apple Silicon generations. Whether that premium translates into workstation-class sales will depend on demand from AI researchers, video professionals and software developers with deep local-compute needs.

Apple's positioning of the Mac Studio as an AI workstation comes as rival chip makers have also emphasized on-device inference in their desktop roadmaps. The M5 Ultra's claimed 4.7-fold AI performance gain over the M3 Ultra, combined with the 512 GB memory ceiling, sets a benchmark that competing workstation vendors will need to address in their next product cycles.

The new Mac Studio and Mac mini are available through Apple's online store and authorized resellers. Apple has not disclosed shipping lead times for the M5 Ultra configuration, which requires assembly of two M5 Max dies in the UltraFusion package and may face constrained supply at launch.