Toloka tests GPT-6 Astra for robot control: 90% success vs 30% for classical stack
MParakhin · x · 2026-11-05
Toloka tested OpenAI's GPT-6 Astra and Typesafe's Jev as robot control policies on a bimanual Trossen Robotics arm.
Key results
- As a visuomotor policy, GPT-6 Astra hit 90% success on a harder pick-and-place test, vs 30% for the classical stack it wrote itself
- Trade-offs: higher latency and roughly $5 per episode
Full comparison is on Toloka's blog. Notable first quantitative data on frontier LLMs acting directly as robot policies (GPT-6 Astra unconfirmed by OpenAI).
More from Embodied
- KAIST's H2RB transfers human demonstrations to robots, boosting object transfer from 8.0 to 51.8 — kaist-ai · 2026-11-05
- New AI glasses UI concept demo shows off futuristic interface design — AnhPhuNguyen1 · 2026-11-05
- Berkeley project uses pretrained base policy's support to guide exploration for fast online adaptation — abhishekunique7 · 2026-11-05
- "Robotics simulation platform" startups are a dead end — agents already removed the need — carlosdponx · 2026-11-05
- Runtime (YC F26) debuts simulation-to-real robot policies, threads a rope with no fine-tuning — ycombinator · 2026-11-05
- Rooms3d turns Vision Pro room scans into permanent 3D dioramas for $9.99, one-time purchase — Scobleizer · 2026-11-05