The Year in AI Papers
The year/Independent research

Paper 2607.13125

Boogu-Image-0.1: Boosting Open Agentic Multimodal Generation via Understanding under a Minimal Budget

Guoxuan Chen, Chufeng Xiao, Haoran Yang, Siyue Xie, Binxiao Huang, Ming Zhang, Cheuk Him Chau, Xinyu Fu, Yingzhao Lian, Tom S. Y. Li, Jintao Lin, Bowen Dong, et al.

Published
Jul 2026
Research lab
Independent
Citations
0
GitHub
Not linked

01 In brief

Summary

Boogu-Image-0.1 is an open-source family of unified multimodal understanding and generation models (Base, Turbo, Edit, Edit-Turbo) that achieves competitive performance in text-to-image generation, fast inference, instruction-based editing, and bilingual text rendering.

The authors argue that strengthening understanding—via a stronger multimodal encoder (Qwen3-VL-8B), agentic prompt rewriting, and related techniques—substantially enhances generation and editing even under constrained compute.

The base model was trained from scratch on only 208.62 million unique images at a theoretical cost of about $400K.

On Boogu Arena, Boogu models lead the open-source tier, trailing only closed-source GPT-Image-2 and Nano-Banana-Pro; on Qwen-Image-Bench, Boogu-Image-0.1-Base-Thinking achieves the best open-source overall scores (53.57 Chinese, 53.73 English).

On ImgEdit-Bench, Boogu-Image-0.1-Edit-Thinking attains the best overall score (4.64).

The paper critiques saturated benchmarks like GenEval and DPG-Bench, which fail to track human preference, and discusses data curation principles, the importance of inference-time cost, and cultural bias in open-source data.

Weights, code, and recipes are released under Apache 2.0.

Limitations include gaps in world knowledge, anatomical inconsistencies, and reliance on the FLUX.1 VAE for fine details.

Future directions include more transparent evaluation, overcoming the open-source data bottleneck, and stronger agentic generation.

Code: https://github.com/Boogu-Project/Boogu-Image.

02 From the paper

Abstract

We introduce Boogu-Image-0.1, an open-source unified multimodal understanding and generation model family, comprising Base, Turbo, Edit, and Edit-Turbo variants. It delivers competitive performance in high-quality text-to-image generation, fast inference, instruction-based editing, and bilingual (Chinese-English) text rendering. Closed-source multimodal systems like Nano-Banana-Pro and GPT-Image-2 achieve strong performance through system-level integration rather than a single model, yet their internal practices remain largely undisclosed. In this work, we demonstrate that strengthening the understanding capability of the system, through a stronger multimodal encoder, agentic prompt rewriting, and related techniques, together with improvements in data quality, training pipelines, and agentic inference-time scaling, can substantially enhance generation and editing performance even under highly constrained compute budgets. Comprehensive evaluations show that Boogu-Image-0.1 consistently matches or surpasses other open-source models across standard benchmarks, and achieves results approaching leading closed-source systems. Notably, this is accomplished with only 208.62 million unique images. The base model's theoretical training cost is only approximately \$400K. We share practical discussions that we believe are valuable to the broader research community, and release weights, code, and recipes under Apache 2.0 to advance the open ecosystem for unified multimodal understanding and generation. Our code is available here: https://github.com/Boogu-Project/Boogu-Image.