Apple has released updated versions of its Mac Studio and Mac mini desktops, featuring new M6 and M5 Ultra chips designed to enhance local AI inference and software development. These updates focus on improving performance for tasks like running large language models, with the M5 Ultra offering up to 512GB of unified memory. The changes reflect a growing emphasis on local AI processing, which has gained traction since macOS 26.2 introduced features supporting distributed AI inference.
The M6 chip, Apple’s first 2nm SoC in its M-series lineup, includes 12 CPU cores, with two designated as super cores, and a 12-core GPU. Apple claims it is up to 40 percent faster at multi-threaded CPU tasks than the M4. It also offers faster unified memory with up to 160GB/s bandwidth, though memory capacity is capped at 32GB. The M5 Ultra, available in the updated Mac Studio, combines two M5 Max chips into a single SoC, delivering 36 CPU cores and 80 GPU cores, along with 1.2TB/s of unified memory bandwidth.
The updates are part of a broader trend toward local AI inference, driven by the popularity of open-weight models like Qwen and DeepSeek. These models allow developers to run large language models locally without relying on expensive cloud services. However, most consumer-grade hardware, such as the average MacBook Pro, still lacks the capacity to handle the largest models, making devices like the Mac Studio and Mac mini valuable for developers seeking to run more complex workflows.
Source: arstechnica
