OpenAI's GPT-6 Astra Achieves 95% Success in Physical Manipulation Benchmark
- Published
- Source
- HumanoidsDaily
OpenAI's GPT-6 Astra has achieved a remarkable 95% success rate in a gross pick-and-place task using dual I2RT YAM robotic arms, according to evaluation data from RoboCurve. This performance highlights Astra's advancements in spatial reasoning and physical manipulation, surpassing competitors like Anthropic's Claude Fable models.
However, the trials also revealed limitations in Astra's zero-shot physical reasoning capabilities, particularly in precision tasks requiring sub-millimeter tolerances. In a precision puzzle task, Astra only succeeded in 2 out of 20 attempts, indicating that while it excels in certain areas, it still faces challenges similar to earlier models.
Looking ahead, RoboCurve's Jay Chooi suggests that if current performance trends continue, large language models like Astra could potentially control robotic arms in real time within the next two to three years. This development could significantly impact the robotics industry, particularly in applications requiring high precision and efficiency.
Original report
OpenAI’s GPT-6 Astra Hits 95% on Physical Manipulation Benchmark, But Precision Tasks Expose Familiar Bottlenecks
HumanoidsDaily · [email protected] (Humanoids Daily Staff)This briefing is an independently written summary based on publicly available reporting and is provided for industry information and news discovery. The original report and source publication are credited and linked where applicable. RobotToday does not claim ownership of third-party source material.
Rights concerns? Contact [email protected] with the relevant URL and details. We will review the matter and take appropriate action where warranted.
More related news
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
- 11
- 12
Related Suppliers
OpenAI
United States
US AI research lab and product company founded 2015; develops GPT, DALL-E, and robotics/embodied AI foundation models targeting physical-world AGI.
Robocurve
United States
San Francisco public benefit corporation founded 2026 that benchmarks frontier AI models on real robot arms and maintains the Inspect Robots framework.
Anthropic
United States
AI safety and research company and public benefit corporation that builds the Claude family of large language models.
Shenzhen Liangyuan Innovation Technology Co., Ltd. 深圳亮源新创科技有限公司
China
Shenzhen embodied-AI lab led by ex-OpenAI researcher Roger Jiang, building Physical AI foundation models Light-O1, Light REACT and LightNav-0
Versatile Robotics
Germany
Wessling, Germany DLR spinoff founded 2024; builds patented MPTC controller software for safe, high-speed human-robot collaboration.
DeepAware AI (Robotics Center)
United States
YC S25 San Francisco start-up supplying physical-AI teams with 90+ robot platforms, teleoperation data collection, a robot data marketplace and the RC M1.
Perceptron, Inc.
United States
Bellevue foundation-model lab founded by two ex-Meta FAIR scientists; its open-weight Isaac models give robots video understanding, reasoning and control.
MABI Robotic AG
Switzerland
Swiss manufacturer of high-precision CNC robots and 6-axis robotic arms for subtractive and additive manufacturing.
Robotic Box Co., Ltd.
Korea, Republic of (South)
Korean AI-powered manufacturing automation specialist offering safety management solutions for forklifts and robotic arms.
Eureka Robotics
Singapore
AI-powered robotic software for high-accuracy precision tasks in manufacturing and logistics with 0.2mm repeatability.
Flexiv Robotics 非夕
United States
Santa Clara adaptive robotics company founded 2016; makes force-controlled Rizon arm series with 0.005N force sensing; raised $322M total.
Humanoid M.AX Alliance 휴머노이드 M.AX 얼라이언스
Korea, Republic of (South)
Seoul-based national humanoid robotics consortium launched April 2025 under South Korea's MOTIE, uniting 224 member organizations.
Robotics needs a service framework.
RSF defines a common language for robot service capability, lifecycle operations, certification pathways, and service-provider networks.
Join the RobotToday community on LinkedIn
Daily robotics news, in-depth analysis, conference highlights, and discussions with professionals worldwide.
