Can MiniMax H3 Run in ComfyUI without Using the Official API?

Overview

Not yet. Running MiniMax H3 in ComfyUI without the official API would require local open weights, and as of early August 2026 those have not been released — the model was announced on July 31, 2026 as open-weight, but the weights have not shipped. So any ComfyUI use today still goes through the hosted API. Source note: Verified 2026-08-06 against the MiniMax official H3 blog, MiniMax Video Generation API docs, and Hugging Face MiniMax-H3 model page.

Once the weights land and a compatible ComfyUI node exists, a fully local, no-API workflow becomes possible: you would load the weights, run generation on your own GPU with enough VRAM, and pay no per-call fee — subject to the license and your hardware. That is a future capability, not a current one.

For now, if you want to use H3 inside ComfyUI, an API node is the way: the interface is ComfyUI, but generation runs on MiniMax’s infrastructure. It is a middle path, not true local generation.

Verify the weight release and node availability before planning a local ComfyUI setup, and re-check the license for local weights when the repository appears, since running locally does not by itself grant commercial rights.