GPT-6 Astra scores 95% on robot control task, 2.3x cheaper than Fable 5.1's 40%

max_paperclips · x · 2026-09-06

A quoted thread claims GPT-6 Astra hit 95% on a robot control task, up from Fable 5.1's 40%, while using 6.2x fewer output tokens at 2.3x lower cost. The reposter mocks the recent European discourse that LLMs were 'saturated' and robotics should leapfrog to non-LLM approaches.

Related event: GPT-6 Astra scores 95% on robot control task at 43% of prior cost(5 posts)→

Original post →

More from Embodied

Embodied channel →