SenseTime Launches Open-Source SenseNova U1 Model Optimized for Chinese Chips
Key Points
- SenseTime released the open-source image model SenseNova U1 on Tuesday.
- U1 processes images directly, eliminating the need to translate visuals into text.
- The model runs on Chinese-made chips; ten domestic chipmakers announced support at launch.
- NEO-Unify architecture underpins faster image generation and lower power use.
- SenseNova U1 is available for free on Hugging Face and GitHub.
- Co‑founder Dahua Lin cites robotics as a key application for the model's speed.
- The release aims to offset U.S. export restrictions that have limited access to advanced AI hardware.
Chinese AI firm SenseTime unveiled SenseNova U1, an open-source image model that processes visuals directly and runs on domestically produced chips. The release aims to offset U.S. export restrictions that have limited the company's access to advanced hardware. Available on Hugging Face and GitHub, the model promises faster image generation and interpretation, with support from ten Chinese chipmakers, including Cambricon and Biren Technology. Co‑founder and chief scientist Dahua Lin said the new architecture could help the firm regain ground in the global AI race and accelerate robotics applications.
Sensing a widening gap with U.S. rivals, SenseTime rolled out SenseNova U1 on Tuesday, offering it to the public for free on Hugging Face and GitHub. The open‑source model, the company says, can generate and interpret images without first converting them to text, a shortcut that cuts processing time and lowers the hardware burden.
"The model's entire reasoning process is no longer limited to text. It can reason with images as well," said Dahua Lin, SenseTime's co‑founder and chief scientist, in an interview. Lin, who also teaches information engineering at the Chinese University of Hong Kong, highlighted the advantage for robots that must make rapid decisions in cluttered environments.
SenseNova U1 runs on Chinese‑made chips, a strategic choice after U.S. export controls barred the firm from accessing the most advanced AI processors, chiefly those from Nvidia. Ten domestic chip designers, among them Cambricon and Biren Technology, announced compatibility with the model on launch day. Lin noted that while the company will continue to explore training on a variety of chips, the best performance may still require top‑tier hardware.
The model's speed stems from a new architecture dubbed NEO‑Unify, which SenseTime previewed earlier this year. According to the company's technical report, NEO‑Unify lets the system handle visual data natively, bypassing the text‑translation step that slows many competitors. The result is faster image generation and lower power consumption, allowing the model to run on PCs and smartphones.
Industry observers see the move as a bid to catch up with both domestic startups like DeepSeek and MiniMax and Western leaders such as OpenAI's GPT‑Image‑2.0. While SenseNova U1 reportedly matches the image quality of leading Chinese closed‑source models such as Alibaba's Qwen and ByteDance's Seedream, its chief selling point is speed.
Open‑sourcing the model also opens the door to broader collaboration. Researchers at Hugging Face praised the release, noting that community testing could surface practical challenges early. Lin stressed that feedback from the open‑source community accelerates iteration, a factor he believes now outweighs the closed‑source versus open‑source debate.
Beyond image tasks, SenseTime envisions robotics applications. By processing visual inputs directly, a robot could more quickly identify objects, evaluate complex machinery, and decide on actions without the latency introduced by text‑based reasoning. Lin hinted at ongoing work with ACE Robotics, a startup co‑founded by another SenseTime executive, to embed the model in future humanoid platforms.
The launch comes amid repeated U.S. sanctions accusing SenseTime of enabling surveillance in Xinjiang, allegations the company denies. Those restrictions have limited foreign investment and technology transfers, prompting the firm to double down on domestic resources and open‑source development as a way to sidestep geopolitical hurdles.
With SenseNova U1 now publicly available, the AI community can test its claims, and SenseTime hopes the model will help the company reclaim a leading position in China's fast‑moving AI landscape.