GPT-6.2 Astra tops robot-code test: 2.5B net solving 79% of LIBERO, Python code hits 96.8%

burny_tech · x · 2026-11-18

The Playground team reports GPT-6.2 Astra is the strongest robot-code author they've tested: with no simulator, demos, or web access, it wrote a 2.5B-parameter neural net solving 79% of 500 LIBERO trials, and Python code that solved 96.8% — beating Fable 5.2 (89.2%) and GPT-6 Astra (33.2%). The generational gap in LLM-generated robot control code is striking.

Original post →

More from Embodied

Embodied channel →