Counterfactual: without reasoning models, AI today might just be reaching o3-level

Jsevillamol · x · 2026-09-03

GregHBurnham offers a progress anchor: had the pre-reasoning scaling trend continued, models today might only now be around o3 / Gemini 2.5 Pro level — or at best Gemini 3 Pro / GPT-5.2 — implying the reasoning paradigm dramatically accelerated progress. He caveats that counterfactual interventions are unknowable, but the estimate frames how much the paradigm shift bought us.

Original post →

More from AGI Musings

AGI Musings channel →