← CanIRunLocal · All hardware · All models
Yes. Llama 3.2 3B fits in fast memory on Mac Studio M4 Max (40c GPU) 128GB at Q8_0 (4.2 GB needed, 96.0 GB usable), about 128 tok/s (estimate).
Hardware: Mac Studio M4 Max (40c GPU) 128GB · Model: Llama 3.2 3B
Mac Studio M4 Max (40c GPU) 128GB has 128 GB of unified memory and 546 GB/s of memory bandwidth. M4 Max (40c GPU) 128GB has the same memory and bandwidth, so the fit numbers match; this page is Mac Studio M4 Max (40c GPU) 128GB. The checker leaves macOS a share of unified memory, so Ollama can use 96.0 GB. Llama 3.2 3B (3.21B, Meta) at Q8_0 is a 3.4 GB download and needs 4.2 GB with headroom for about 8k tokens.
| Best quant | Q8_0 |
|---|---|
| Download size | 3.4 GB |
| Memory needed | 4.2 GB |
| Usable fast memory | 96.0 GB |
| Fit | fits in fast memory |
| Estimated speed | ~128 tok/s (estimate) |
| Command | ollama run llama3.2:3b-instruct-q8_0 |
Estimates, not benchmarks. Model sizes are ollama.com download sizes; memory needed adds headroom for ~8k tokens of context; speeds are rough figures from memory bandwidth. Data updated 2026-10-23.
A project by OnDevice Labs · llms.txt