GPT-6-Astra Evaluated Across 12 Robot Benchmarks; Reasoning Latency Emerges as the Bottleneck
ccloy · x · 2026-10-19
SenseNova Team Releases Embodied Intelligence Evaluation of GPT-6-Astra
A joint research team from SenseNova (60 authors) published "Has GPT-6-Astra Achieved Embodied Intelligence?" on arXiv, probing GPT-6-Astra's robotic manipulation and navigation competence:
- Scope: Large-scale evaluation across 12 representative benchmarks spanning tabletop manipulation, humanoids, mobile platforms, and dexterous hands.
- Unified protocol: Allows in-context learning (ICL) from training-set demonstrations and substantially increases execution steps. Under this protocol, GPT-6-Astra achieves unprecedented performance, challenging the conventional paradigm of building specialized models per task and embodiment.
Core Claim: Reasoning Latency Is Part of Embodied Competence
The authors argue that on physical robots, retries consume time, incur wear, and can cause irreversible failures. In their prototype household benchmark, the world keeps running during inference — an agent may correctly decide to turn off the stove yet act too late to prevent an egg from burning.
This motivates benchmarks grounded in human needs, without in-domain demonstrations, and with a clock that keeps ticking. Technical report, project page, and code are all released.
Related event: GPT-6-Astra Embodied Evaluation: Strong Benchmarks, Real-World Gaps(4 posts)→
More from Embodied
- MolmoAct2 Lead Duan Jiafei Joins NUS, Wins Multimillion-Dollar Embodied AI Grant — DJiafei · 2026-10-19
- Sipeed's MaixStation challenges DGX Spark: 128GB LPDDR5X at 40W, one-third the size — AIFlow_ML · 2026-10-19
- A Virtual Fly Driven by Its Real 70k-Neuron Connectome Runs Live 24/7 on Twitch — WatchInternational89 · 2026-10-19
- Robot brain inference cost fell 84% in two months, from $2.36 to $0.38 per task — victor_explore · 2026-10-19
- NORINCO unveils Fuxi, an anti-riot humanoid robot that mirrors its operator's movements in real time — CyberRobooo · 2026-10-19
- HexaShell: VLM Coding Agents Drive Dual-Arm Robots, SOTA on RoboDojo — HexaFuture · 2026-10-19