GPT-5.6 Leads on the DeepSWE Leaderboard

koltregaskes · x · 2026-07-11

GPT-5.6 Sol max and xhigh have surpassed Fable 5's xhigh and max on the DeepSWE leaderboard, averaging lower costs, fewer output tokens, and fewer steps.

Additionally, GPT-5.6 Tera max scored almost identically to Fable.

Related event: GPT-5.6 Tops DeepSWE Leaderboard with Superior Cost-Efficiency(11 posts)→

Original post →

More from Models

Models channel →