Daxiao Robotics' 9B ACE-Brain-0.6 Beats GPT-6 Astra on Spatial Grounding

Daxiao Robotics has released ACE-Brain-0.6, a 9B-parameter embodied foundation model built to improve visual spatial anchoring. The company trained the model on about one million high-quality, verifiable spatial-anchoring samples, aiming to help embodied systems connect language instructions with objects and locations in physical space.

In eight Grounding evaluations, ACE-Brain-0.6 outperformed OpenAI's GPT-6 Astra on six tasks. On seven public benchmarks, it averaged 74.53 points, compared with 73.68 for GPT-6 Astra—a 0.85-point edge. Daxiao Robotics says the results show that targeted spatial-anchoring data can let a smaller model compete with larger general-purpose systems on embodied AI tasks.

The release places spatial grounding at the center of robot learning: before a robot can follow a command, it must reliably identify the relevant object, region, or position. Daxiao Robotics frames ACE-Brain-0.6 as progress on that capability, while the reported comparisons are limited to public grounding benchmarks rather than full robotic deployments.