Boogu-Image-0.1 is an open-source family of models, including Base, Turbo, Edit, and Edit-Turbo variants. It supports text-to-image generation, fast inference, instruction-based editing, and bilingual text rendering.
The work demonstrates that strengthening understanding capability through a stronger multimodal encoder, agentic prompt rewriting, improved data quality, and refined training pipelines can substantially enhance generation and editing performance even under highly constrained compute budgets.
Comprehensive evaluations show Boogu-Image-0.1 consistently matches or surpasses other open-source models across standard benchmarks and approaches leading closed-source systems. This is accomplished with only 208.62 million unique images and a theoretical training cost of approximately $400K.