Nodes/MiniMax H3 Combined Model Loader/MiniMax H3 Combined Model Loader
ComfyUI Node

MiniMax H3 Combined Model Loader

Load MiniMax H3 as FL2VA+REF2VA combined on the fly, or load either FL2VA or REF2VA as a normal standalone diffusion model.

By entropicnoise·Created 11 days ago·Updated 9 days ago· 3
MiniMax H3 Combined Model Loader
    • model
    fl2va_model
    ref2va_model
    ref_start_block25
    ref_end_block49
    weight_dtypedefault
    attentiondefault
    allow_compilefalse
    load_modeload fl2va+ref2va
    Categorymodel/loaders

    Inputs (8)

    NameTypeDefaultDescription
    fl2va_modelCOMBOFL2VA checkpoint. Used by 'load fl2va+ref2va' and 'load fl2va model'.
    ref2va_modelCOMBOREF2VA checkpoint. Used by 'load fl2va+ref2va' and 'load ref2va model'.
    ref_start_blockINT250–49Combined mode only: first block whose adaln_proj comes from REF2VA (inclusive).
    ref_end_blockINT490–49Combined mode only: last block whose adaln_proj comes from REF2VA (inclusive).
    weight_dtypeCOMBOdefaultSame weight dtype override as ComfyUI's Load Diffusion Model node. Leave at default for embedded quantizations such as INT8 ConvRot.
    attentionCOMBOdefaultAttention backend. 'default' leaves ComfyUI unchanged; 'pytorch' forces the PyTorch/SDPA path. SageAttention modes require their corresponding optional package. SageAttention3 requires supported Blackwell hardware.
    allow_compileBOOLEANfalseAllow torch.compile around SageAttention extension calls. Leave disabled unless your SageAttention setup supports it.
    load_modeCOMBOload fl2va+ref2vaCombined mode uses both files and the selected REF2VA block range. Standalone modes ignore the range.

    Outputs (1)

    NameTypeDescription
    modelMODEL