The Hong Kong-founded firm said the multimodal system combines language and vision in one model and can deliver image-generation results at one-tenth the cost of leading international rivals.
SenseTime is targeting enterprise customers as Chinese AI groups race to cut costs, win users and find sustainable revenue, while larger platform companies such as Alibaba, Tencent and ByteDance retain funding advantages.
The company, under US sanctions it denies are justified, said overseas expansion remains focused on Southeast and North Asia, the Middle East and Brazil after narrowing its net loss 58.6% last year.
As China's state-backed AI undercuts global prices, can America's tech sector win a price war it isn't built for?
Is China's government-fueled AI boom a sustainable global strategy or just an enormous, fast-burning cash bubble?
Are US sanctions unintentionally forging a more efficient AI rival by forcing Chinese firms to innovate past hardware limits?
On April 28, 2026, SenseTime open-sourced the SenseNova U1 series under the Apache 2.0 license, making advanced multimodal AI models freely available for commercial use and self-hosting. Powered by the innovative NEO-Unify architecture introduced earlier that year, these models unify pixel-level image processing and language understanding in a single framework, eliminating traditional visual encoders and reducing latency by 30%. This breakthrough enables strong performance in image generation, logical reasoning, and robotics applications, while simplifying deployment and lowering costs. The open-source strategy fosters a global developer ecosystem, circumvents sanctions, and accelerates adoption, setting the stage for future larger models and progress toward artificial general intelligence.