Apple announces new Mac Studio and Mac mini models

Apple announced Mac Studio models with M5 Max and M5 Ultra, alongside Mac mini models with M6 and M5 Pro.[1][2] Apple says the M5 Ultra Mac Studio can be configured with up to 512GB of unified memory, which it says is enough to run enormous language models entirely on the device.[1] The M5 Max configuration has an 18-core CPU, up to a 40-core GPU and up to 128GB of unified memory.[1] Apple claims the new Studio delivers up to 4.3 times faster AI performance, while the M6 mini delivers up to four times faster AI performance.[1][2]

The new Mac mini is available for preorder starting today, with availability beginning September 22.[2] Ars Technica reported that preorders for both product lines start today and that they ship September 22, except for the 512GB Mac Studio configuration, which is due in late October.[3] Ars Technica reported that the M6 Mac mini starts at $899 with 16GB of memory, while Mac mini configurations with M5 Pro start at $1,699.[3] The M5 Max Mac Studio starts at $2,499, and M5 Ultra configurations start at $5,499, according to Ars Technica.[3]

Apple and Meta revealed a collaboration that enables developers to run Meta’s open-weight AI model on Apple’s high-end desktop computers.[4] Meta describes Muse Spark Open as an 816-billion-parameter model, with 42 billion active parameters and a one-million-token context window.[5] Meta says it can run on a desktop using two M5 Ultra Mac Studios and the MLX framework.[5] Meta also says the model was built for coding and agentic work, including long-horizon tasks and complex software engineering.[5] Its weights are released under a license with limited commercial terms that allow many developers to use the model for free.[5]

Apple says distributed AI inference can be up to three times faster than on a single system.[1] Ars Technica reported that developers have also been linking Mac minis and Mac Studios to run local models too large for a single mass-market device.[3] Meta says Muse Spark Open is available through API platforms including Baseten, Fireworks AI, Groq, Modal, Nebius, OpenRouter and Together AI.[5] Apple and Meta said local operation adds a layer of security and saves users money compared with cloud use.[4] Those statements describe the companies’ claims; the evidence does not provide independent performance, security or cost comparisons.