LLM serving, synthesized for your Mac.
Up to 2.7× as fast as the best other server on the same Mac.
See it run
On a 48 GB M4 Pro Mac: install the Flint engine, serve it, and run Claude Code on it with jit-launch claude. The agent finds and fixes two bugs in a small project, on the local model. Waits are fast-forwarded, as the status bar says.
Get started