GPT-6 hits 95% on robot control; joker asks if it just API-calls MolmoAct2

chris_j_paxton · x · 2026-09-06

A quoted tweet claims GPT-6 Astra scored 95% on a robot control task, up from Fable 5.1's 40%, with 6.2x fewer output tokens at 2.3x lower cost. DJiafei jokes: could GPT-6 just API-call the MolmoAct2 checkpoint and ask it to put the block into the bowl? — implying the score may lean on an existing embodied-policy model underneath.

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

Original post →

More from Fun

Fun channel →