Apple

Mac mini (M5 Pro, 2026)

M5Released September 22, 2026Launch price $1,699 (Apple US price, 15-core CPU, 16-core GPU, 24GB / 512GB)

The pro Mac mini: an M5 Pro with up to 18 CPU and 20 GPU cores, Thunderbolt 5 and twice the M6's memory bandwidth (307GB/s), which is what sets the speed of a local language model. Its 24GB base memory runs 7-14B models; configured with 48GB or 64GB it holds 4-bit models in the 30B class and up, in a box the size of a hardback. For local AI it is the cheapest current Mac with Pro-class bandwidth; for gaming its GPU sits well behind the Mac Studio's.

Buy now Mac mini 55days since update
Just updated: the next model is a long way off. Last updated August 25, 2026, Mac mini (M6 and M5 Pro); the last 4 updates came every 714 days on average.

Against the line-up

Geekbench 6 CPU Single-core · points

#2 of 3 · The Verge; † another publication
  1. 1Mac mini M64,708
  2. 2Mac mini M5 Pro4,294†
  3. 3Mac mini M43,838

Geekbench 6 CPU Multi-core · points

#1 of 4 · The Verge; † another publication
  1. 1Mac mini M5 Pro29,104†
  2. 2Mac mini M4 Pro22,661
  3. 3Mac mini M621,405
  4. 4Mac mini M414,838

Geekbench 6 GPU Metal · points

#1 of 1 · The Verge
  1. 1Mac mini M5 Pro142,708

Blender Benchmark samples/min

#1 of 4 · Blender Open Data
  1. 1Mac mini M5 Pro3,643
  2. 2Mac mini M62,656
  3. 3Mac mini M4 Pro2,570
  4. 4Mac mini M41,276

The whole Mac mini line-up side by side, with prices

Every benchmark

Quoted from the publications that ran them; we don't test hardware ourselves. Figures from different publications were run on different systems and scenes, so compare within a row's source.

Local LLM (llama.cpp / Ollama)

SettingResultSource
Llama 2 7B Q4_0, Generation
one source
67.6tok/s llama.cpp community benchmarks (GitHub), Oct 19, 2026
M5 Pro, 16-core GPU, 307GB/s
llama-bench tg128 (generating 128 tokens), all layers on the GPU.
Llama 2 7B Q4_0, Prompt processing
one source
1,340tok/s llama.cpp community benchmarks (GitHub), Oct 19, 2026
M5 Pro, 16-core GPU, 307GB/s
llama-bench pp512 (reading a 512-token prompt), all layers on the GPU.
Llama 2 7B Q8_0, Generation
one source
38.9tok/s llama.cpp community benchmarks (GitHub), Oct 19, 2026
M5 Pro, 16-core GPU, 307GB/s
llama-bench tg128 (generating 128 tokens), all layers on the GPU.

Blender Benchmark

SettingResultSource
Standard run
one source
3,643samples/min Blender Open Data, Oct 19, 2026
Apple M5 Pro (GPU - 16 cores), Metal, Blender 5.2.0
Median score over the three scenes, 14 submitted runs
CPU
one source
500samples/min Blender Open Data, Oct 19, 2026
Apple M5 Pro CPU, Blender 5.2.0
Median score over the three scenes, 44 submitted runs. Blender Open Data does not split the 15- and 18-core CPUs.

Geekbench 6 CPU

SettingResultSource
Multi-core
one source
29,104points The Verge, Sep 21, 2026
Mac mini M5 Pro, CTO 18-core CPU, 20-core GPU, 48GB, 1TB
Single-core
one source
4,294points The Verge, Sep 21, 2026
Mac mini M5 Pro, CTO 18-core CPU, 20-core GPU, 48GB, 1TB

Geekbench 6 GPU

SettingResultSource
Metal
one source
142,708points The Verge, Sep 21, 2026
Mac mini M5 Pro, CTO 18-core CPU, 20-core GPU, 48GB, 1TB
OpenCL
one source
88,531points The Verge, Sep 21, 2026
Mac mini M5 Pro, CTO 18-core CPU, 20-core GPU, 48GB, 1TB

Sources

As an Amazon Associate we earn from qualifying purchases, and we may earn from other shop links. Amazon's price is on Amazon; Best Buy's is read from Best Buy and shown with the time it was checked.