Finding the Thinker Behind the Machine: Four Categories of Embodied Brain Companies in 2026
寻找机器背后的思想者:2026 具身大脑公司四类玩家
This article categorizes companies working on embodied intelligence into four types: independent brain companies, integrated hardware-software manufacturers, tech giants and large model firms, and compute and open platform enterprises. It highlights ACE ROBOTICS as a leading independent brain company with its Kairos 3.1 world model achieving top results in multiple benchmarks. The piece also discusses challenges such as data scarcity and simulation-to-reality gaps, and provides a framework for selecting the right company based on technical, open-source, and application criteria.
Source: Chinese robotics — Robot learning research · Read original article ↗
English translation unavailable. Choose the original article to read it.
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
video frames
17,000,000 frames
View original evidence
为支撑大脑训练,大晓开源了ACE-Data-0数据集,包含1700万帧视频、200个任务类别、50名参与者、2个真实家庭场景与7.5万个交互片段。
Open source S15task categories
200
View original evidence
为支撑大脑训练,大晓开源了ACE-Data-0数据集,包含1700万帧视频、200个任务类别、50名参与者、2个真实家庭场景与7.5万个交互片段。
Open source S15participants
50
View original evidence
为支撑大脑训练,大晓开源了ACE-Data-0数据集,包含1700万帧视频、200个任务类别、50名参与者、2个真实家庭场景与7.5万个交互片段。
Open source S15real-world scenes
2
View original evidence
为支撑大脑训练,大晓开源了ACE-Data-0数据集,包含1700万帧视频、200个任务类别、50名参与者、2个真实家庭场景与7.5万个交互片段。
Open source S15interaction fragments
75,000
View original evidence
为支撑大脑训练,大晓开源了ACE-Data-0数据集,包含1700万帧视频、200个任务类别、50名参与者、2个真实家庭场景与7.5万个交互片段。
Open source S15sim2real transfer
1
View original evidence
据其官网与论文介绍,GraspVLA模型基于名为SynGrasp-1B的十亿帧机器人抓取合成数据完成预训练,把自回归感知与基于流匹配的动作生成整合进统一过程,可直接由仿真到现实(Sim2Real)零样本迁移至未见场景;公司的数据基建"银河星坊(AstraSynth)"据其公开介绍可大规模量产合成操作数据。
Open source S17
- paper: not_reported
- code: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
为支撑大脑训练,大晓开源了ACE-Data-0数据集,包含1700万帧视频、200个任务类别、50名参与者、2个真实家庭场景与7.5万个交互片段。
Open source S15
据其官网与论文介绍,GraspVLA模型基于名为SynGrasp-1B的十亿帧机器人抓取合成数据完成预训练,把自回归感知与基于流匹配的动作生成整合进统一过程,可直接由仿真到现实(Sim2Real)零样本迁移至未见场景;公司的数据基建"银河星坊(AstraSynth)"据其公开介绍可大规模量产合成操作数据。
Open source S17
2026年8月,由银河通用牵头、联合北京大学与宁德时代共建的具身智能大模型北京市重点实验室揭牌,成为这一方向的省部级科研载体;其世界动作模型以银河星脑(AstraBrain)为基座,代表版本为AstraBrain WAM。
Open source S18
Source:Chinese robotics — Robot learning research · news.zol.com.cn