GPT-6 Sol Scores 1483 Elo on Knowledge Work, Far Behind Sonnet 5.5's 1811

haider1 · x · 2026-09-29

Blogger haider1 compiled long-horizon knowledge work results that look brutal for GPT-6 Sol:

Combined with his earlier Terminal-Bench findings — Sonnet 5.5 beats Opus 5.5 there and roughly ties it on knowledge work and computer use at half the price — Anthropic appears to have compressed most of the Opus experience into Sonnet this generation, while GPT-6 Sol lags badly on knowledge work.

Original post →

More from Models

Models channel →