ACE Robotics, founded by SenseTime co-founder Xiaogang Wang, introduced its embodied artificial intelligence model, ACE-Brain-0.6, on October 19. This model was developed through fine-tuning Qwen3.5-9B and is designed for object recognition, determining their location in images, and establishing interaction points for robots.
According to the company's statement, the model was trained using approximately one million spatial grounding samples. In conducted evaluations, ACE-Brain-0.6 demonstrated superiority over GPT-6 Astra in six out of eight grounding benchmarks.
The model scored an average of 74.53 across seven publicly available benchmarks, and achieved a result of 94.13 on the RobotGrounding test, which is based on real-world robot scenarios. These figures are higher than those of Astra, which scored 73.68 and 91.10 respectively.