DeepGen 1.0: A Lightweight Unified Multimodal Model for Image Generation and Editing

ABSTRACT

Current unified multimodal models for image generation and editing typically rely on massive parameter scales (>10B), entailing prohibitive costs. We present DeepGen 1.0, a lightweight 5B unified model that achieves comprehensive capabilities competitive with much larger counterparts. We introduce Stacked Channel Bridging (SCB), a deep alignment framework that extracts hierarchical features from multiple VLM layers and fuses them with learnable 'think tokens'. Despite being trained on only ~50M samples, DeepGen 1.0 surpasses the 80B HunyuanImage by 28% on WISE and the 27B Qwen-Image-Edit by 37% on UniREditBench, providing an efficient alternative for multimodal research.