> ## Documentation Index
> Fetch the complete documentation index at: https://docs.comfy.org/llms.txt
> Use this file to discover all available pages before exploring further.

# Model Merge MiniMax H3 - ComfyUI Built-in Node Documentation

> Merge two compatible MiniMax H3 models by component, with a separate ratio for each part of the architecture.

Merge two compatible MiniMax H3 models by component, with a separate ratio for each part of the architecture. Each ratio is the share of `model1`, so you can keep the patch projections and refinement blocks of one model while replacing the transformer blocks with another.

## Inputs

| Parameter | Description | Data Type | Required | Range |
| - | - | - | - | - |
| `model1` | First MiniMax H3 model. Every component starts from it, and each ratio decides how much of that component comes from `model2` instead. | MODEL | Yes | - |
| `model2` | Second MiniMax H3 model that supplies the components you blend in. | MODEL | Yes | - |
| `video_patch_proj.` | Share of `model1` for the video patch projection (default: 1.0). | FLOAT | Yes | 0.0 to 1.0 (step: 0.01) |
| `audio_patch_proj.` | Share of `model1` for the audio patch projection (default: 1.0). | FLOAT | Yes | 0.0 to 1.0 (step: 0.01) |
| `condition_proj.` | Share of `model1` for the conditioning projection (default: 1.0). | FLOAT | Yes | 0.0 to 1.0 (step: 0.01) |
| `time_embedder.` | Share of `model1` for the time embedder (default: 1.0). | FLOAT | Yes | 0.0 to 1.0 (step: 0.01) |
| `token_refiner.blocks.0.` | Share of `model1` for the first token refiner block (default: 1.0). | FLOAT | Yes | 0.0 to 1.0 (step: 0.01) |
| `token_refiner.blocks.1.` | Share of `model1` for the second token refiner block (default: 1.0). | FLOAT | Yes | 0.0 to 1.0 (step: 0.01) |
| `token_refiner.final_norm.` | Share of `model1` for the token refiner final normalization (default: 1.0). | FLOAT | Yes | 0.0 to 1.0 (step: 0.01) |
| `blocks.0.` through `blocks.49.` | Share of `model1` for each of the 50 transformer blocks, one widget per block (default: 1.0). | FLOAT | Yes | 0.0 to 1.0 (step: 0.01) |
| `final_layer.norm.` | Share of `model1` for the final layer normalization (default: 1.0). | FLOAT | Yes | 0.0 to 1.0 (step: 0.01) |
| `final_layer.adaln_proj.` | Share of `model1` for the final adaptive normalization projection (default: 1.0). | FLOAT | Yes | 0.0 to 1.0 (step: 0.01) |
| `final_layer.video_out.` | Share of `model1` for the video output layer (default: 1.0). | FLOAT | Yes | 0.0 to 1.0 (step: 0.01) |
| `final_layer.audio_out.` | Share of `model1` for the audio output layer (default: 1.0). | FLOAT | Yes | 0.0 to 1.0 (step: 0.01) |

A ratio of 1.0 (the default) keeps that component entirely from `model1` and 0.0 takes it entirely from `model2`; values in between blend the two. Tensors that carry no ratio, such as normalization statistics, always come from `model1`. Both inputs must be MiniMax H3 models with the same architecture, otherwise the merge produces a model that does not run.

## Outputs

| Output Name | Description | Data Type |
| - | - | - |
| `MODEL` | The merged MiniMax H3 model. | MODEL |

> This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! [Edit on GitHub](https://github.com/Comfy-Org/embedded-docs/blob/main/comfyui_embedded_docs/docs/ModelMergeMiniMaxH3/en.md)

***

**Source fingerprint (SHA-256):** `4fea4e588d16a8396c68aa071a59e671a2a0c645734606be6d7e221986ba88bd`


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.