Apple has officially announced two new processors — the flagship M5 Ultra and the M6 — which the company has used to equip its refreshed desktop computers, the Mac Studio and Mac mini. Both chips focus on accelerated artificial intelligence workloads, complex graphics, and local processing of large language models. According to the developers, the M5 Ultra is the most powerful processor in the company's history, while the M6 is Apple's first chip built on a 2-nanometer process. Pre-orders for the Mac mini with the M6 are already open, starting at $899, with official shipments beginning after September 22.
The M5 Ultra flagship: two M5 Max crystals in a single chip
The M5 Ultra processor is positioned as the peak of Apple's compute lineup and is intended for professionals working with 3D graphics rendering, visual effects creation, and running modern AI models. The chip's key feature is its architecture: it is the company's first processor created by combining two dual-die M5 Max chips into a single solution. This design achieved a unified memory bandwidth of 1.2 TB/s, which is 50 percent higher than its predecessor, the M3 Ultra. According to Apple, it is precisely this enormous memory bandwidth that unlocks "unprecedented desktop-class power" and allows large models to run without data-transfer bottlenecks.
M6: the first 2-nanometer chip for the mass market
The second announcement — the M6 processor — is aimed at everyday users, developers, and AI enthusiasts. It is built on a 2-nanometer process, which, according to the company, delivers a high level of energy efficiency. The M6 features a 12-core central processing unit (two cores more than the M5), an updated graphics block, and a dual 16-core Neural Engine. Thanks to this configuration, the performance of graphical AI computations has increased by nearly 30 percent compared to its predecessor. Apple describes the M6 as a chip delivering a "revolutionary leap" in both performance and energy efficiency.
Local AI and data privacy
Both processors are designed to run and fine-tune large AI models directly on Mac computers, without relying on cloud computing. Apple's tools automatically distribute the workload between the CPU, the graphics block, and the Neural Engine, making it possible to use both Apple's own Foundation Models framework and third-party models. On-device data processing, the company claims, guarantees a high level of privacy and creates a secure environment for working with users' sensitive data and experimenting with new AI tools.
Contradictory data
When the announcement is compared against source materials, one inconsistency in the interpretation of the lineup emerges, and it is worth sorting out. On the one hand, by generation number, the M6 looks "newer" than the M5 Ultra and is built on a more advanced 2-nanometer process — this is supported by sources calling the M6 the "first 2-nm chip." On the other hand, in terms of actual power and positioning, the M5 Ultra is the true flagship: it is assembled from two pairs of M5 Max crystals, has a memory bandwidth of 1.2 TB/s, and is billed as the most powerful processor in Apple's history, whereas the M6 targets the mass market. In other words, the "higher number" does not mean the "higher class" in this pair: the M5 Ultra and the M6 are two different poles of the lineup (a professional flagship and an energy-efficient mass-market chip), not sequential versions of a single device. It is also worth noting separately that the memory bandwidth comparison is made against the M3 Ultra, not the M5 Max, which can be misleading when reading the figures "at a glance."
Pricing and shipping timelines
The new Mac mini with the M6 processor is already available for pre-order, starting at $899; official shipments of the computers will begin after September 22. According to sources, the top-tier Mac Studio configuration with the M5 Ultra megachip is priced at roughly $18,000, confirming the premium positioning of the flagship desktop solution. Thus, Apple is expanding its desktop lineup in two directions at once — maximum compute power for professionals and energy efficiency for a broad range of users — with a focus on local, private artificial intelligence.
