OpenAI's GPT-6 Astra tops a new ranking of robot AI models
Playground, which says it builds real-world environments to benchmark robots, published the Robot Intelligence Index on Oct. 30, 2026. It combines 16 public sim and real-robot benchmarks into one chess-style rating per model. Among the 30 models with both real-robot and simulation results, OpenAI's GPT-6 Astra rates highest at 1905, ahead of InternW0-Δ (1828), GPT-6.1 Sol (1808), Gemini 4 Argon (1797) and MWA-Phoenix (1774), Playground says.
Astra also leads the 12 models tested on two or more real-robot benchmarks and ranks first on precision tasks. Dream-1 has the highest real-robot rating, from RoboDojo-RC, and GO-3 Nano Preview leads in simulation.
Why it matters
No single robot benchmark tests more than a fraction of today's models, so Playground calls this the broadest comparison yet. Astra stayed first in nearly every reweighting it tried, but the weights are Playground's judgment calls, and real-robot rank ranges are less reliable, it says.
The source post
Post by Playground on X.
View the post on X
Sources
New to robots? Start with our beginner's guide to how robots work.
Summarized by Servospace from the sources above, drafted with AI help and checked against those sources before publishing. Claims are the companies' or outlets' unless we say otherwise. How we work.