MiniMax H3 Max and H3 Max Turbo Join ComfyUI Partner Nodes

ComfyUI's MiniMax video nodes now include H3 Max and H3 Max Turbo, two hosted variants of the open-weight MiniMax H3 model (H3 Max PR, Reference node PR, Max Turbo PR).

H3 Max is a post-trained variant of open H3 tuned for stronger prompt adherence and better aesthetics, positioned as the quality-focused tier. H3 Max Turbo is a further distillation of H3 Max for throughput: the same look at a fraction of the generation time, released as a preview in early September 2026. Both trade the 2K output ceiling of the open base model for faster hosted generation at 480p and 768p, with clips from 5 to 15 seconds and synchronized audio.

What changed in ComfyUI

The MiniMax partner nodes gained new model options across three workflow families:

• Text to Video and First-Last-Frame nodes: H3 Max and H3 Max Turbo as selectable model options (PR #16025, PR #16094).
• Reference to Video node: H3 Max support for reference-driven generation (PR #16041).

The official H3 Max Turbo text-to-video template

The official H3 Max text-to-video template

How to use it

Update ComfyUI to the latest build and open any MiniMax H3 template or workflow: the H3 Max and H3 Max Turbo options appear in the model dropdown of the text-to-video, first-last-frame, and reference nodes. Generation runs on MiniMax's hosted infrastructure through the API node system with billing per output second, and the open-weight H3 checkpoints remain fully available for local generation.

These variants slot alongside the existing acceleration ecosystem covered in the H3 Acceleration Arena: where community distills like Fused Turbo and FastH3 optimize local inference, H3 Max and Max Turbo provide the hosted quality and speed tiers.