H3Shot DeskRun planner Search/
All H3 reference notes
Deployment

How to deploy MiniMax H3 privately or locally

Answer

MiniMax H3 can run locally for 768P with the open H3-Base-FL2VA and H3-Base-Ref2VA checkpoints. Use SGLang, vLLM, Diffusers, or ComfyUI. This is not full hosted parity: H3-Context-IR and H3-Regenerate-2K are not in the open release, so the official full 2K workflow still crosses an API boundary.

Try this sequence
  1. Choose FL2VA for text-only or first/last-frame jobs, and Ref2VA for multi-reference image, video, or audio jobs.
  2. Download only the checkpoint family and components your serving framework needs from the official H3 repository.
  3. Start with the official SGLang, vLLM, Diffusers, or ComfyUI recipe and validate a reproducible 768P output.
  4. Before calling the setup private, document GPU memory, license terms, stored inputs, and whether Context-IR or 2K regeneration uses a hosted API.
Read local deployment guideOpen MiniMax sourceVerified against MiniMax docs / Aug 5, 2026