8B MoT
32K tokens
Apache 2.0
Open-weight native multimodal SenseNova release built to unify multimodal understanding, reasoning, and generation inside a single monolithic architecture rather than using separate visual encoders or VAEs. The model card presents SenseNova U1 as a NEO-Unify-based system that models language and visual information end to end as a unified compound, supports image understanding, image generation, image editing, and interleaved image-text generation, and is intended as a step toward natively multimodal agentic learning. The currently public open release consists of two 8B Lite checkpoints: a base 8B-MoT model and an 8B-MoT-SFT model. Release materials claim open-source state-of-the-art performance across multimodal understanding and generation benchmarks, strong dense information rendering for posters, presentations, comics, resumes, and other infographic-style outputs, and native interleaved generation for illustrated tutorials and visually grounded long-form content. The card also notes a current visual understanding context limit of 32K tokens and describes the SFT models as trained through understanding warmup, generation pre-training, unified mid-training, unified SFT, and an initial round of text-to-image RL training.