Vals AI launches Science Index: GPT-6 Astra tops 21 models, no model sweeps every discipline

teortaxesTex · x · 2026-10-17

Vals AI released its Science Index measuring models' ability to make scientific discoveries across 21 models. GPT-6 Astra ranks #1 overall, followed by GPT-6.1 Sol, Claude Sonnet 5.5, Claude Opus 5.5, and Gemini 4 Argon. No single model leads every science: Sonnet 5.5 takes #1 in BioMysteryBench, Physical Sciences and Engineering Sciences (but falls to #11 in Earth Sciences), while GPT-6 Astra dominates MysteryMechanism, Terminal-Bench Science, Life Sciences and Math. Commenters note Anthropic's RL progress is moving fast given Sonnet 5.5's three individual wins.

Related event: Vals AI Science Index Crowns GPT-6 Astra, No Model Dominates All Fields(3 posts)→

Original post →

More from Models

Models channel →