On October 29, Daxiao Robotics unveiled its own embodied spatial intelligence interface, ACE-Brain-0.6API, putting the ability to "understand space" directly into developers' hands. This interface can tap three core capabilities: recognizing what target is in front of it, precisely localizing that target in 3D space, and then determining where people and objects can interact—essentially giving machines a set of "spatial perception nerves" that can understand the physical environment.
What made this "brain" smarter was the expansion of its training data. Daxiao Robotics disclosed that the model's spatial anchoring training data was fed from about 1.1 million entries all the way to 1.5 million, equivalent to adding nearly 40% more material to its spatial memory. After this thicker data foundation, the model's average score climbed to 75.41 across seven public Grounding benchmarks, not only firmly holding above its previous level, but also surpassing OpenAI's GPT-6Astra. For a domestic model aiming at embodied intelligence, being able to directly outperform a frontier flagship on the foundational track of spatial understanding means it is one step closer to truly directing robots to see clearly and move to the right places.