llama.cpp on the M2 Pro 32GB
On the M2 Pro 32GB, llama.cpp runs Qwen3.8 27B at 4-bit at a median 8.6 tokens per second for generation, from 1 plain run (128K context).
Speed by model on the M2 Pro 32GB
Plain runs only: one device, one request, no speculative decoding, the whole model in memory. Each row is one model size at one bit class, with the other engines' plain runs of the same model on this card beside it. Context length and build still differ between runs, and each side shows its own.
| Model | llama.cpp | Other engines |
|---|---|---|
| Qwen3.8 27B 4-bit | 8.6 t/sone report, 128K context | no other engine reported |
3 llama.cpp reports on the M2 Pro 32GB
Newest first. A run on several cards, with speculative decoding, with part of the model in system RAM or with batched requests says so, since each describes a different setup.
- Model not identified · Q4_K_M33.6 t/s
- Model not identified · Q4_K_M34.8 t/s
- Qwen3.8 27B · IQ4_XS8.6 t/s
Frequently asked
How fast is llama.cpp on the M2 Pro 32GB?
On the M2 Pro 32GB, llama.cpp runs Qwen3.8 27B at 4-bit at a median 8.6 tokens per second for generation, from 1 plain run (128K context).