GPT-6 Astra on robot arms | Robocurve
OpenAI's GPT-6 Astra vs Claude Fable 5.1 controlling a pair of YAM arms under the same agent policy: 19/20 vs 8/20 on block-into-bowl in interleaved blinded pairs, 2/20 vs 2/20 on the puzzle, with 80% fewer output tokens.
September 4, 2026 A follow‑up to our comparison of Claude Fable 5 and Fable 5.1. We gave OpenAI's GPT‑6 Astra control of the same YAM arms under the same Inspect Robots agent policy, on the same two tasks: “Pick up the red block from the table and place it inside the bowl.” “Pick up the round blue puzzle piece by the knob at its center and place it into the matching circular groove in the board.” On the bowl task Astra placed the block in 19 of 20 trials, against Fable 5.1's 8 of 20 and Fable 5 in 1 of 20, in 2.5 minutes per trial to Fable 5.1's 6.8, at an estimated $0.94 per run to…
saved by
related reading
- GPT-6 Astra System Carddeploymentsafety.openai.com
- OpenAI Jalapeño: Better Than Nvidia Blackwellnewsletter.semianalysis.com
- How Claude Performs on Robotics Tasks \ Anthropicanthropic.com
- GPT-6 Astra System Card - OpenAI Deployment Safety Hubdeploymentsafety.openai.com
- Astra for Coding: Why Are We Doing This Again?lucumr.pocoo.org
- GPT-6 Astra System Card - OpenAI Deployment Safety Hubdeploymentsafety.openai.com
- Robocurve — Real-world evaluations of physical AIrobocurve.org
- CAIS AI Dashboarddashboard.safe.ai
- Fable is SOTA at CIFAR Speedrun (& specification gaming): lessons on AI R&D automationfulcrum.inc
- gpt-4.pdfcdn.openai.com
- Astra Is Hard to Monitorthezvi.substack.com
- TERMINAL-BENCHtbench.ai