LLM serving, synthesized for your Mac.
Up to 2.7× as fast as the best other server we measured on the same Mac.
We used SkyDiscover to build specialized engines for a 24 GB M2 and a 48 GB M4 Pro. Each engine targets a specific model and Mac configuration, but we benchmarked all of them on both machines.
Set your Mac and memory, then choose a model below to get its commands.
Choose your model
- 2.72×Qwen3.8-Flash-Next on a 48 GB M4 Pro, against slotstream
- 1.36×Qwen3.6 on a 24 GB M2, against llama.cpp with MTP
- 1.22×Flint on a 48 GB M4 Pro at 252K tokens of context, against rec