SOTA open-source instruction-based image editing model using MLLM + DiT architecture. Comparable to GPT-4o and Gemini 2 Flash on editing tasks.

Paper

arXiv: 2504.17761

visioneditinggenerationopen-weight

Notes

arXiv submission Apr 24, 2025. Step1X-Edit-v1p2 (ReasonEdit) released Nov 2025.