Mac mini M6
A desktop mini computer from Apple featuring the M6 chip and a unified memory architecture. The initial release offers a 24 GB memory configuration, making it suitable for everyday productivity workloads and lightweight AI inference tasks.
Timeline (last 90 days)
To judge whether a 32 GB Mac can run a given model, one should refer to macOS's peak memory footprint (39.0 GB for ERNIE-Image Turbo), not RSS (which reported only 5.1 GB in the test).
The Mac mini M6 24 GB version cannot launch the Qwen 3.8B model because only about 10 GB of memory is available, and entering 'Hello' immediately triggers an out-of-memory error.
Apple's M6 Mac mini launched with only the 24 GB version available; the 32 GB version has not yet been released by the company.
The macOS 27 system on a Mac mini M6 uses approximately 12 GB of memory at startup, and in actual use can reach 14 GB.
Users who want to run local large language models on a Mac mini should give up on the 24 GB version and wait at least for the 32 GB version.
At a price of around 1,000+ USD, the Mac mini achieves a local large language model inference speed of approximately 10 tokens per second.
The 32 GB version of the Mac mini M6 is not expected to be available until mid-October.
All Facts (5)
The macOS 27 system on a Mac mini M6 uses approximately 12 GB of memory at startup, and in actual use can reach 14 GB.
60%UnverifiedTo judge whether a 32 GB Mac can run a given model, one should refer to macOS's peak memory footprint (39.0 GB for ERNIE-Image Turbo), not RSS (which reported only 5.1 GB in the test).
50%UnverifiedThe Mac mini M6 24 GB version cannot launch the Qwen 3.8B model because only about 10 GB of memory is available, and entering 'Hello' immediately triggers an out-of-memory error.
50%UnverifiedUsers who want to run local large language models on a Mac mini should give up on the 24 GB version and wait at least for the 32 GB version.
50%UnverifiedAt a price of around 1,000+ USD, the Mac mini achieves a local large language model inference speed of approximately 10 tokens per second.
50%