GPT 5.6 Sol Takes the Lead on WeirdML

teortaxesTex · x · 2026-07-10

Following the WeirdML v2 update, GPT 5.6 Sol (high) scored 88.8% on the WeirdML benchmark, slightly edging out Claude Fable 5 at less than half the price. The author also noted Sol's remarkable consistency, never dropping below 60% of the task's SOTA score across 85 separate runs.

Original post →

More from Models

Models channel →