KTransformers

MiniMax Models

MiniMax support is currently inference-focused.

Inference

ModelEntryMethodStatus
MiniMax M2kt run m2FP8Current / Needs smoke.
MiniMax M2.1kt run m2.1FP8Current / Needs smoke.
MiniMax M2.5Manual SGLang-KT tutorialFP8Manual / Needs smoke; not registry-first yet.

Documentation Source

Current material should be migrated from:

  • doc/en/kt-kernel/MiniMax-M2.1-Tutorial.md
  • doc/en/MiniMax-M2.5.md

Before publishing performance wording, record the exact model path, --kt-method, hardware tuple, and smoke prompt.