HasGPT-6-AstraAchievedEmbodiedIntelligence？
Overview
NVIDIA's 'Astranomy' research investigates the use of GPT-6 Astra for robot control without additional training, achieving a 74.8% success rate on the RoboLab-120 benchmark. The study evaluates three control methods—Direct, Agentic, and Orchestrated—demonstrating the impact of AI agent history management on performance. Challenges include inference speed, API costs, and stability in object manipulation.
From Robotics — Robot learning news discovery
Event evidence and corrections
0 attributed source owners. Ownership does not establish independent confirmation. Quantities are reported separately and are never added together.
Reported quantity · Reported duration: 8 hours · Basis not reported · Differing source assertions
“在此次大会上，宇树科技正式发布了续航长达 8 小时的 H2-X 功夫传奇人形机，并宣布推进物理 AI 自进化体系与 世界模型 生态落地。”
Exact source · revision 2Source owner not reported
Reported quantity · Reported duration: 3 hours · Basis not reported · Differing source assertions
“大会新一代产品 最长续航约8小时 （前代约3小时），增强了长时间值守的现实性。”
Exact source · revision 2Source owner not reported
Reported quantity · Reported completion rate: 95 percent · Basis not reported
“模型侧变化更具标志意义： 新一代通用大模型已可直接读取相机流与本体状态、输出末端位姿完成真机任务 ， 标准化抓取测试完成率95%、单次成本约1美元 ，并 显著压缩token用量 。”
Exact source · revision 2Source owner not reported
Reported quantity · cost: 1 other · Basis not reported
“模型侧变化更具标志意义： 新一代通用大模型已可直接读取相机流与本体状态、输出末端位姿完成真机任务 ， 标准化抓取测试完成率95%、单次成本约1美元 ，并 显著压缩token用量 。”
Exact source · revision 2Source owner not reported
Reported quantity · Reported success rate: 100 percent · Basis not reported · Differing source assertions
“》对GPT-6Astra的具身能力进行测试，通过特定的条件（In-contextlearning+最高10倍步数预算）测试，发现在长程机器人操作基准LIBERO-Long++评测上成功率达100%（此前SOTA82%），操作鲁棒性和泛化能力的LIBERO-Pro上成功率99.3%（此前SOTA50%），人形机器人基准SIMPLE成功率达98.8%（此前SOTA92.5%），灵巧操作DexJoCo成功率87.3%（此前SOTA40.3%）。”
Exact source · revision 2Source owner not reported
Reported quantity · Reported success rate: 99.3 percent · Basis not reported · Differing source assertions
“》对GPT-6Astra的具身能力进行测试，通过特定的条件（In-contextlearning+最高10倍步数预算）测试，发现在长程机器人操作基准LIBERO-Long++评测上成功率达100%（此前SOTA82%），操作鲁棒性和泛化能力的LIBERO-Pro上成功率99.3%（此前SOTA50%），人形机器人基准SIMPLE成功率达98.8%（此前SOTA92.5%），灵巧操作DexJoCo成功率87.3%（此前SOTA40.3%）。”
Exact source · revision 2Source owner not reported
Reported quantity · Reported success rate: 98.8 percent · Basis not reported · Differing source assertions
“》对GPT-6Astra的具身能力进行测试，通过特定的条件（In-contextlearning+最高10倍步数预算）测试，发现在长程机器人操作基准LIBERO-Long++评测上成功率达100%（此前SOTA82%），操作鲁棒性和泛化能力的LIBERO-Pro上成功率99.3%（此前SOTA50%），人形机器人基准SIMPLE成功率达98.8%（此前SOTA92.5%），灵巧操作DexJoCo成功率87.3%（此前SOTA40.3%）。”
Exact source · revision 2Source owner not reported
Reported quantity · Reported success rate: 87.3 percent · Basis not reported · Differing source assertions
“》对GPT-6Astra的具身能力进行测试，通过特定的条件（In-contextlearning+最高10倍步数预算）测试，发现在长程机器人操作基准LIBERO-Long++评测上成功率达100%（此前SOTA82%），操作鲁棒性和泛化能力的LIBERO-Pro上成功率99.3%（此前SOTA50%），人形机器人基准SIMPLE成功率达98.8%（此前SOTA92.5%），灵巧操作DexJoCo成功率87.3%（此前SOTA40.3%）。”
Exact source · revision 2Source owner not reported
Reported quantity · Reported success rate: 95 percent · Basis not reported · Reported zero-shot result; the excerpt does not specify an evaluation protocol. · Differing source assertions
“破壳机器人创始人许华哲表示，GPT-6Astra在机器人控制任务中成功率达95%，前代模型Fable5.1只有40%，在RoboLab语义理解任务中，Astra直接控制的零样本成功率为98%。”
Exact source · revision 2Source owner not reported
Reported quantity · Reported success rate: 98 percent · Basis not reported · Reported zero-shot result; the excerpt does not specify an evaluation protocol. · Differing source assertions
“破壳机器人创始人许华哲表示，GPT-6Astra在机器人控制任务中成功率达95%，前代模型Fable5.1只有40%，在RoboLab语义理解任务中，Astra直接控制的零样本成功率为98%。”
Exact source · revision 2Source owner not reported
Reported quantity · Reported success rate: 74.8 percent · Basis not reported · Differing source assertions
“Overview of NVIDIA 'Astranomy' - Achieving a 74.8% Success Rate in Robot Control with GPT-6 Astra”
Exact source · revision 2Source owner not reported
Reported quantity · Reported success rate: 74.8 percent · Basis not reported · Differing source assertions
“ranomy provides GPT-6 Astra with camera images, robot state, and task instructions to generate actions directly. In the robot manipulation benchmark ' RoboLab-120 ', it achieved a 74.8% success rate , outperforming the comparative robot-specific models. The main features are as follows: ・Robot control without additional training Directly generating robot actions with GPT-6 Astra ・Three types of control methods Compar”
Exact source · revision 2Source owner not reported
Evidence dependency changes (13)
- 2026-10-30T21:33:18.536Z · evidence updated · source revision 2. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-30T21:33:01.232Z · evidence updated · source revision 2. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-30T21:23:22.737Z · evidence updated · source revision 2. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-30T20:44:34.651Z · evidence updated · source revision 2. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-30T20:44:04.603Z · evidence updated · source revision 2. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-30T20:37:02.342Z · evidence updated · source revision 2. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-30T20:36:38.879Z · evidence updated · source revision 2. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-30T20:29:59.033Z · evidence updated · source revision 2. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-30T20:29:32.307Z · evidence updated · source revision 2. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-30T20:08:50.433Z · evidence updated · source revision 2. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-30T20:08:41.040Z · evidence updated · source revision 2. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-23T03:05:07.176Z · evidence updated · source revision 2. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-23T01:37:09.912Z · evidence updated · source revision 2. Evidence extraction was updated; current source attributions are shown above.
Developments
- 2026-10-25 13:32 UTC · 1 reportsHasGPT-6-AstraAchievedEmbodiedIntelligence？Chinese robotics — Dataset and collection discovery：GPT-6Astra Embodied Intelligence Breakthrough and Physical AI Data Flywheel Initiation
- 2026-10-22 14:51 UTC · 1 reportsOpenAI GPT-6 Astra model release and embodied intelligence roadmapChinese robotics — Robot learning research：GPT-6 Astra and the Future of Embodied Intelligence
- 2026-10-22 00:15 UTC · 1 reportsIn 2027, Not Letting General Models Control Robots Isn't Cutting-Edge AIChinese robotics — Research news discovery：In 2027, Not Letting General Models Control Robots Isn't Cutting-Edge AI
- 2026-10-20 14:37 UTC · 1 reportsAstranomy: Robot Control with GPT-6 AstraRobotics — Robot learning news discovery：Overview of NVIDIA 'Astranomy' - Achieving a 74.8% Success Rate in Robot Control with GPT-6 Astra
Report timeline
Follow attributed reports and material updates.
- Chinese robotics — Dataset and collection discoveryGPT-6Astra Embodied Intelligence Breakthrough and Physical AI Data Flywheel Initiation
A research paper evaluates GPT-6Astra's performance in embodied intelligence tasks, achieving high success rates on benchmarks like LIBERO-Long++, SIMPLE, and DexJoCo. The study also explores the implications of these results for physical AI self-evolution and the growing data demands in the field.
- Chinese robotics — Robot learning researchSignalPhysical AI and Embodied Intelligence: Industry Signals Point Toward a Critical Inflection Point
The article discusses the convergence of physical AI and embodied intelligence, with key industry events and product launches signaling a shift toward standardized robot configurations and model-driven production. The release of H2-X and GPT-6Astra demonstrates progress in direct model control and physical interaction, while policy support and market data suggest the industry is approaching a critical mass.
- Chinese robotics — Robot learning researchSignalGPT-6 Astra and the Future of Embodied Intelligence
The article explores how large language models such as GPT-6 Astra are entering the field of embodied intelligence, with implications for both the industry and research. It discusses the capabilities and limitations of these models, the potential for collaboration with specialized embodied models, and the challenges in real-world deployment. The piece also examines the impact on investment and valuation in the sector, as well as the evolving role of data and simulation in training these models.
- Chinese robotics — Research news discoveryIn 2027, Not Letting General Models Control Robots Isn't Cutting-Edge AI
The article explores the evolving role of large language models in embodied AI, focusing on how models like GPT-6 Astra and Opus-5.5 are being tested for robotic control. It discusses the challenges and potential of using general-purpose models for physical tasks, including performance metrics and limitations such as inference latency and token cost. The piece also touches on industry shifts and the implications for startups and research directions in the field.
- Robotics — Robot learning news discoveryOverview of NVIDIA 'Astranomy' - Achieving a 74.8% Success Rate in Robot Control with GPT-6 Astra
NVIDIA's 'Astranomy' research investigates the use of GPT-6 Astra for robot control without additional training, achieving a 74.8% success rate on the RoboLab-120 benchmark. The study evaluates three control methods—Direct, Agentic, and Orchestrated—demonstrating the impact of AI agent history management on performance. Challenges include inference speed, API costs, and stability in object manipulation.
Timezone · UTC
Article dates follow your selected timezone. Briefing editions use Hong Kong time (UTC+8).