Apple unveils M6 (first 2nm Mac chip) and quad-die M5 Ultra to boost on‑device AI

Apple is pitching its latest Mac chips as an architectural leap as much as a routine speed bump: The company said Monday it has introduced its first 2-nanometer Mac chip, the M6, and its first quad-die Ultra chip, the M5 Ultra, with both debuting in new desktop Macs. In a press release published Aug. 25, Apple said it “debuted M6 in the new Mac mini and M5 Ultra in the new Mac Studio.”

The significance, Apple says, is not just faster Macs but more local computing power for artificial intelligence tasks that run on the device instead of in the cloud. For now, though, the performance figures come from Apple’s own testing on preproduction systems in August 2026, and there are no independent third-party benchmarks yet. Apple’s testing note says the results “reflect the approximate performance” of the new Mac mini and Mac Studio.

Apple describes the M6 as “Apple’s first 2-nanometer chip.” It said the chip includes a 12-core CPU with 2 super cores, 4 performance cores and 6 efficiency cores, along with a 12-core GPU, a dual 16-core Neural Engine, support for up to 32 gigabytes of unified memory and up to 170 GB/s of memory bandwidth. Apple said the M6 delivers the “world’s fastest single-threaded performance,” up to 1.2 times faster multithreaded performance than M5 and up to 2.4 times faster than M1. For AI workloads, Apple said the M6 offers nearly 30% higher peak GPU compute than M5 and more than eight times that of M1.

The M5 Ultra is the more unusual design. Apple said it is the company’s first quad-die architecture, built by using UltraFusion to connect two dual-die M5 Max chips so they behave as a single system on a chip. Apple said the M5 Ultra can be configured with up to a 36-core CPU, including 12 super cores and 24 performance cores, and up to an 80-core GPU. It supports up to 512 GB of unified memory and 1.2 TB/s of memory bandwidth, which Apple said is 50% more than M3 Ultra. Apple said the chip delivers up to 1.25 times higher single-threaded performance than M3 Ultra, up to 1.3 times higher multithreaded performance and up to 4.5 times the peak GPU compute for AI. Apple also said the M5 Ultra includes a more capable Media Engine with dedicated H.264 and HEVC, four ProRes encode and decode engines, and hardware-accelerated AV1 decode.

The quad-die design stands out because Apple first introduced UltraFusion with the M1 Ultra in 2022 as a way to link two dies into one chip. Apple says M5 Ultra extends that idea further. The company is also tying both chips to a broader push toward on-device AI, saying developer tools including Core AI, Core ML, Metal and Xcode will use the chips for local model execution, and that developers can run and fine-tune large AI models directly on a Mac. Apple also linked that work to Apple Intelligence features slated for macOS 27 this fall.

“Today, we’re debuting the next giant leap in performance and AI compute for Apple silicon with the incredibly advanced M6 and the most powerful M-series chip yet, M5 Ultra,” Sri Santhanam, Apple’s vice president of Silicon Engineering Group, said in the release.