Blogger: OpenAI models crush Claude Opus at scientific automation, not just benchmarks

VraserX · x · 2026-09-30

Blogger VraserX argues OpenAI models dominate math, science, computer use and 3D/world modeling, showing they're built for serious scientific work rather than coding benchmarks. He concedes Claude Opus is better at coding, but claims even GPT-6.1 Sol, a smaller distilled model, crushes it at automating real science — a gap Tech Twitter underappreciates.

Original post →

More from Models

Models channel →