Apple has announced new Mac Studio and Mac Mini offerings which the firm has promoted as options to be used for running AI locally.
On Tuesday, the tech giant formally announced the M5 Max and M5 Ultra Mac Studio as well as the M5 Pro and M6 Mac Mini, which the firm says offer a generational leap over their predecessors and are optimised for AI inference.
Apple said the Mac Studio is “the ultimate desktop for on‑device AI”, delivering up to 4.3x faster AI performance than the M3 Ultra Mac Studio and a 50 per cent leap in memory bandwidth to 1.2TB/s.
Likewise, Apple said its new Mac Mini offerings are up to 4x faster at AI tasks than their predecessors and specifically highlighted the devices as useful for running local models and powering “always-on agentic computing”.
Developer teams looking to run larger and more powerful AI models on Mac can connect multiple Mac Studios together via Thunderbolt 5 or remote direct memory access (RDMA), to form a cluster that Apple says will run frontier open-weight models.
Using the 512GB memory configuration of the M5 Ultra Mac Studio, users could run models such as Meta’s Llama 4 70B, mid-sized variants of Alibaba’s Qwen 3 Coder, and DeepSeek R1.
Johny Srouji, chief hardware officer at Apple, said: “With the powerful M5 Max and the incredible capabilities of M5 Ultra, Mac Studio ushers in a new era of desktop computing, delivering huge performance gains for pro workloads and AI inference with frontier-class models.
“By integrating Neural Accelerators directly into the GPU and offering massive amounts of high-bandwidth unified memory, the new Mac Studio is our most powerful Mac ever.”
Running AI models locally allows users to complete tasks such as code generation for free, and without data sharing concerns. The practice became a major trend towards the end of 2025, with the launch of the open-source AI agent OpenClaw.
Running on the energy efficient and compact Mac Mini, OpenClaw allowed users to run 24/7 background AI tasks with low up-front and running costs. M6 Mac Mini
The Mac Studio competes directly with the Nvidia’s DGX Spark, an AI-focused compact desktop computer powered by a Grace Blackwell CPU and a GB10 Superchip. Clustered together, four DGX Spark devices offer a similar performance to the M5 Ultra Mac Studio but are more optimised for AI, with the majority of LLMs having been written specifically for Nvidia’s CUDA computing platform.


.jpg)


