Apple Mac mini (M6)

It suits developers and everyday users who work in macOS and want a compact computer. At 127 x 127 x 50.8 mm with an internal power supply, it needs no brick on the desk. For local AI, check the exact model release, quantization, context length, runtime overhead and memory available after the operating system. System RAM and GPU VRAM are different limits; shared-memory allocation depends on firmware, drivers and the backend. CPU inference or partial offloading can change what loads and its speed. Installed capacity alone does not establish compatibility or useful generation speed. Check the current Ollama or LM Studio hardware documentation for the exact GPU and operating system.

Skip it if you plan to add memory or drives later, because there are no memory slots, no M.2 slots and no SATA bays. It is a poor fit for workstation duty, and macOS limits hypervisor choice for a homelab. There is no VESA mount for hanging it behind a monitor.

Everyday use is well suited: the twelve-core M6 runs macOS with Wi-Fi 7 and three Thunderbolt ports.

Coding is well suited too, with twelve cores, unified memory from 16 GB to 32 GB and support for three displays. The memory cannot be upgraded later, so pick the size you need at purchase; the 24 GB and 32 GB models leave more room than the 16 GB ones.

A homelab is workable: you get 2.5GbE as standard with a 10GbE option, but memory and storage are fixed at purchase and macOS limits which hypervisors you can run.

Local AI is workable. The 16 GB and 24 GB models share their unified memory between CPU and GPU, which suits small to mid-size quantized models. For local AI, check the exact model release, quantization, context length, runtime overhead and memory available after the operating system. System RAM and GPU VRAM are different limits; shared-memory allocation depends on firmware, drivers and the backend. CPU inference or partial offloading can change what loads and its speed. Installed capacity alone does not establish compatibility or useful generation speed. Check the current Ollama or LM Studio hardware documentation for the exact GPU and operating system. Apple does not state the memory bandwidth, and no generation speed figure is available.

Workstation use is a poor fit: there is no discrete graphics, no x8 or x16 card slot, no memory slots and too little memory for that work.

All four configurations share the twelve-core Apple M6 and its 12-core GPU; only memory and storage change. The 16 GB model with 256 GB of storage is the entry point for everyday work, which is well suited. The 16 GB model with 512 GB gives you more room with the same memory. The 24 GB model with 512 GB adds memory for coding, and that memory cannot be added later. For local AI, check the exact model release, quantization, context length, runtime overhead and memory available after the operating system. System RAM and GPU VRAM are different limits; shared-memory allocation depends on firmware, drivers and the backend. CPU inference or partial offloading can change what loads and its speed. Installed capacity alone does not establish compatibility or useful generation speed. Check the current Ollama or LM Studio hardware documentation for the exact GPU and operating system.

There is no upgrade path inside this machine. Memory is soldered, there are no memory slots, and the maximum is 32 GB, which the top configuration already ships with. Storage works the same way: with no M.2 slots and no SATA bays, the internal drive you buy is the one you keep. Any extra space has to come from external drives instead. Apple does not state ECC support or memory speed.

Apple states support for up to three displays, and the single HDMI output is the only dedicated video port. Apple does not state the HDMI version. You get three Thunderbolt ports and no USB-A ports, so USB-A devices need an adapter or hub. The USB-C port count is also three. Wired networking is one Ethernet port at 2.5GbE with a 10GbE option, and wireless is Wi-Fi 7 with Bluetooth 6. For high-speed external connections, the Thunderbolt ports are what you have; there is no OCuLink.

Memory is soldered and capped at 32 GB, and with no M.2 slots or SATA bays, nothing inside can grow. There are no USB-A ports, no SD card reader and no VESA mount. Apple does not state the memory bandwidth, memory speed, HDMI version or TDP. The base model's 256 GB of storage cannot be expanded internally.

If the memory ceiling decides it for you, three machines go higher. The Lenovo ThinkCentre M70q Gen 5 uses an Intel Core i5-13400T and takes up to 64 GB rather than 32 GB. The MINISFORUM MS-A2 pairs an AMD Ryzen 9 9955HX with a maximum of 96 GB. The GMKtec NucBox K11 uses an AMD Ryzen 9 8945HS and also reaches 96 GB. Each swaps the Apple M6 for an Intel or AMD processor, which is the other difference to weigh.