Apple introduced updated versions of its Mac Studio and Mac Mini desktops on Tuesday, positioning them as platforms for local artificial intelligence development and inference. The company highlighted the new M6 and M5 Ultra chips within these machines, which offer substantial unified memory capacity and bandwidth. These hardware enhancements are designed to support the execution of large language models directly on the device, providing an alternative to cloud-based solutions or specialized GPU hardware.
The Mac Mini now ships with the M6 chip, or an M5 Pro option for more demanding workloads. Apple states the M6 chip includes a 12-core CPU, a 12-core GPU, and a dual 16-core Neural Engine, claiming up to four times faster AI performance compared to the previous M4 model. The M6 Mac Mini can be configured with up to 32GB of unified memory, while the M5 Pro option supports up to 64GB. Both Mac Mini variants incorporate Wi-Fi 7, Bluetooth 6, and 2.5Gb Ethernet as standard, with a 10Gb Ethernet option available.
For more intensive professional tasks, the Mac Studio integrates the M5 Max and the new M5 Ultra chips. The M5 Max configuration features an 18-core CPU, up to a 40-core GPU with Neural Accelerators, and up to 128GB of unified memory. The M5 Ultra scales up to a 36-core CPU, an 80-core GPU, and a maximum of 512GB of unified memory, which Apple states enables users to run large language models entirely on the device. Apple claims the Mac Studio with M5 Ultra delivers up to 4.3 times faster AI performance and up to 1.8 times faster graphics compared to its predecessor. This model also features Wi-Fi 7, Bluetooth 6, and Thunderbolt 5 connectivity.
A key aspect of Apple's strategy involves the ability to link multiple Mac Studio systems for distributed AI inference. Utilizing Thunderbolt 5, which offers low-latency communication between hosts, developers can create a shared memory pool across several machines. Apple suggests that a cluster of four Mac Studio systems can achieve up to three times the AI inference performance of a single unit. This capability builds on a trend that emerged after macOS 26.2 shipped last December, which enabled low-latency communication for distributed AI inference using MLX, an open-source array framework designed for Apple silicon's unified memory. This approach allows hobbyists and professional developers to chain Mac Minis or Mac Studios together to run larger local language models than a single device could handle, offering an alternative to specialized hardware with Nvidia GPUs.
Apple is also providing developer tools and frameworks to support these local AI capabilities. The company's Core AI technologies are designed for on-device AI, assisting developers in deploying full-scale large language models on Apple Silicon. The Core AI framework, along with Core AI Optimization and Core AI PyTorch Extensions, helps prepare and convert models for Apple silicon. Additionally, the Foundation Models framework, a native Swift API, grants direct access to Apple Foundation Models and supports other language model providers. These tools allow developers to run and fine-tune large AI models locally on their Mac, with automatic optimization across the CPU, GPU, and Neural Engine.
The pricing for the new Mac Mini starts at $899, while the Mac Studio with the M5 Ultra begins at $5,499. Pre-orders for the new Mac Mini opened on August 25, with availability slated for September 22.