SenseTime, the Chinese AI company known for its facial recognition technology, released a new open source multimodal model in April 2026 that it says can generate and interpret images faster than competing models from US and Chinese rivals.
The model, called SenseNova-U1, is available for free on Hugging Face and GitHub. Its key technical feature is the ability to process images directly without first converting them to text — a method SenseTime says reduces computing power requirements and speeds up output. The approach is built on a technical structure the company calls NEO-Unify. “The model’s entire reasoning process is no longer limited to text. It can reason with images as well,” said Dahua Lin, cofounder and chief scientist at SenseTime, in an interview with WIRED.
U1 is designed to run on Chinese-made chips. On release day, ten Chinese chip designers — including Cambricon and Biren Technology — announced hardware support for the model. That compatibility is significant given that US export controls restrict Chinese companies from accessing advanced AI chips primarily manufactured by Western firms such as Nvidia.
In its accompanying technical report, SenseTime claims U1 produces higher-quality images than other open source models currently available, with performance comparable to closed source Chinese models from Alibaba and ByteDance. The company acknowledges, however, that it still trails industry leaders like OpenAI’s GPT-Image-2.0.
SenseTime’s pivot to open source comes as the company has struggled to keep pace with newer Chinese AI startups like DeepSeek and MiniMax following the rise of natural language processing systems. Lin said the company shifted its strategy last year after finding that open source development accelerated iteration through researcher feedback.
The release also carries geopolitical context. SenseTime has been sanctioned multiple times by the US government over allegations that its facial recognition technology was used in surveillance systems targeting Uyghurs and other minority groups in China’s Xinjiang region — allegations the company denies. Those sanctions restrict US firms from investing in SenseTime or selling it certain technologies without a license. Going open source, Lin suggested, allows the company to continue working with international researchers outside those constraints.
Source: WIRED