i1: A Simple and Fully Open Recipe for Strong Text-to-Image Models

2026-06-09

AI Summary

This work addresses the significant performance gap between existing fully open-source text-to-image diffusion models and state-of-the-art closed-source counterparts, which is exacerbated by opaque training data and insufficient reproducibility details. Through over 300 controlled experiments—accumulating more than 700,000 TPU v6e hours—the study systematically investigates modeling and data design choices, introducing several efficient strategies, including a balanced mixture of curated datasets and lightweight adapter modules for large text encoders. Leveraging only publicly available data, the authors train i1, a 3-billion-parameter open-source model that surpasses the current best open-source models by an average of 29.5 percentage points across five benchmarks, including GenEval. This represents the first fully open-source model to approach the performance of leading closed-source systems, with weights, code, and data processing pipelines comprehensively released.

Abstract

Diffusion models have consistently driven progress in text-to-image generation. However, it is challenging to attribute recent progress to specific modeling and data choices: state-of-the-art open-weight models provide limited ablations, and do not disclose their training data and full training details. The research community needs fully open (weights, data, and code) models as a foundation for further research; yet existing fully open models still fall significantly short of leading models in performance. In this project, we conduct a systematic investigation of the modeling and data design choices in text-to-image diffusion training and inference with 300+ controlled experiments totaling 700K+ TPU v6e hours. Our experiments highlight several empirical findings (e.g., equal weighting is a strong default for mixing curated datasets) and simple design decisions (e.g., larger text encoder adapters improve performance with minimal added parameters) for training strong models. Guided by these insights, we train i1, a 3B-parameter text-to-image diffusion model using only publicly available datasets. i1 is competitive with leading models on five representative benchmarks (GenEval, DPG, PRISM, CVTG-2K, and LongText), and outperforms the best existing fully open model by 29.5 absolute percentage points on average. We provide the i1 checkpoints, training and inference code, and the data processing pipeline. Together, our findings and the i1 recipe establish a practical foundation for future open research in text-to-image diffusion models. Our code is available at https://github.com/zlab-princeton/i1.

Problem

• text-to-image generation
• diffusion models
• open-weight models
• training data transparency
• reproducible research

Innovation

• open-weight diffusion models
• systematic ablation study
• text-to-image generation
• public dataset training
• efficient adapter design