What is the largest LLM the Apple M5 (16GB) can run?

Mistral Nemo 12B, at Q5_K_M using about 10.6 GB of the 10.7 GB available, at roughly 11 tokens per second (8K context).

How many local AI models run well on the Apple M5 (16GB)?

21 of the 69 models we track run fully in its 16 GB of unified memory at Q4_K_M or better.

Generation speed follows memory bandwidth. With 153 GB/s, expect about 11 tokens per second on an 8B model.

Every model we track, checked against the Apple M5 (16GB) at 8K context. The largest that runs well is Mistral Nemo 12B.

Unified memory16 GB10.7 GB usable by the GPU

Bandwidth153GB/s

Runs well21/69models at Q4_K_M or better

Fastest model~68tok/s · gpt-oss 20B

Bars show memory used out of 10.7 GB. Pick a row for the full breakdown.

NameMemory usedBest quantSpeed

NameMemory usedBest quantSpeed

NameMemory usedBest quantSpeed

Run it in the cloud

Rent a cloud GPU by the hour. No upfront cost, stop any time.