ComfyUI Wiki · 工作流与开源创作分享·· 9 days agoSame storyAI score67
MiniMax releases open-source image model MiniMax-H3-Image, supporting editing with up to 16 reference images.
Text to image
AI briefing
On 2026-11-10, MiniMax released the open-weight image model MiniMax-H3-Image. A single checkpoint supports both text-to-image and editing with up to 16 reference images, with approximately 16 MP output and 6-step sampling without CFG. The model is based on a 33.1B DiT and the Qwen3-VL-32B text encoder; text generation is a core selling point. On the same day, ComfyUI v0.45.0 added native support, with official T2I and Edit templates built in.
Open original comfyui-wiki.com
Same story, the pick is “MiniMax open-sources the H3-Image image model, supporting native 4K and up to 16 reference images”