Apple Announces M6 and M5 Ultra: A Leap Forward in AI Computing

On August 25th, 2026, Apple introduced the M6 and M5 Ultra. The M6 is used in the new Mac mini while the M5 Ultra is found inside the Mac Studio. Apple is promoting these computers as local AI inference devices.

1M6: Apple's First 2nm Chip

M6 is the first M-series chip to use a smaller process node (2nm). It contains a 12-core CPU and a 12-core GPU with Neural Accelerators. The Neural Engine is also doubled in size to 16 cores compared to the M5’s 8 cores. Apple claims that the M6 provides a 1.2x improvement in multi-threaded performance compared to the M5 and is 2.4x faster than the M1. There’s also an improvement of almost 30% in AI compute on the GPU compared to the M5. Unified memory bandwidth is 170 GB/s and can hold up to 32 GB of memory.

2M5 Ultra: 80-Core GPU and 512 GB of Memory

M5 Ultra is the first quad-die chip in the M-series. It is a quad-chip design with four chips connected together in one package. This chip is fully loaded with a 36-core CPU (12 super-cores and 24 performance-cores), an 80-core GPU with Neural Accelerators, and a 32-core Neural Engine. For AI applications, it has 1.2 TB/s of memory bandwidth (50% more than M3 Ultra) and up to 512 GB of unified memory. Its AI performance on GPUs is 4.5x better than the M3 Ultra.

With 512 GB of unified memory, you can run GPT-4 models without quantization. Now, local AI inference on your Mac becomes a viable option for developers and researchers who want privacy and the ability to work offline without needing access to cloud GPUs.