HiDream-I1: 17B Open Source Text to Image Model for ComfyUI

HiDream-I1 is a 17B parameter text-to-image model by HiDream-ai, combining DiT with MoE architecture. MIT licensed, available in Full/Dev/Fast variants.

H

HiDream-I1 by HiDream-ai

Text-to-ImageOpen SourceDiTMoE

Open-source 17B parameter text-to-image model combining DiT with MoE (Mixture of Experts) architecture. MIT licensed. Integrates four text encoders (OpenCLIP ViT-bigG, CLIP ViT-L, T5, LLaMA 3.1 8B). Available in Full (50 steps), Dev (28 steps), and Fast (16 steps) variants, with fp8/GGUF/nf4 community versions.

DeveloperHiDream-ai
Release Date2025-04-07
ArchitectureDiT + MoE (17B parameters)
LicenseMIT
Text EncodersOpenCLIP ViT-bigG, CLIP ViT-L, T5XXL, LLaMA 3.1 8B
VariantsFull (50 steps), Dev (28 steps), Fast (16 steps)
Community Variantsfp8, GGUF (city96), nf4 (azaneko)