KittenML Open-Sources inferencekit Plugin to Optimize Local Hardware Inference

On November 13, 2026, at 08:14, a post appeared on the LocalLLaMA subreddit reporting that KittenML has open-sourced a free plugin called inferencekit. The plugin is designed to turn coding agents such as Claude Code, Codex, and OpenCode into inference engineers for the hardware a user actually runs, helping them automatically locate the optimal inference path for their local setup.

According to the post, most models ship with inference settings tuned for whatever hardware their creators developed on, which typically does not match the hardware a given user has locally. The fastest inference path on a 3070 may differ from the fastest path on a Mac or on a Radeon card, and inferencekit is intended to find that path for each specific configuration.

The post also states that the tool has already been used to bring the streaming TTFB of Kitten-TTS-2 down to roughly 10ms, though the description in the material is cut off at the hardware reference, so the exact environment involved is not fully known. As of now, only this single report has surfaced, and no further developments on the release have been documented.