AI's Ability to Predict Experiments Is Surging, Tests Show

Independent tests by Periodic Labs show models' ability to predict experimental outcomes has improved sharply since GPT-4 Turbo, with Opus 5.5 showing 2.3x the research taste of top human experts at roughly 1/30 the cost per experiment.

2026-10-07 ~ 2026-10-07 · 2 related posts

• Opus 5.5 shows 2.3x the research taste of top human experts at ~1/30th the cost — SaxenaNayan · 2026-10-07
• Models rapidly improve at predicting experiment outcomes, may close 90% of gap by 2030 — SaxenaNayan · 2026-10-07