Model Preferences Have Flipped

MatthewBerman · x · 2026-07-11

In a cited conversation, the author mentioned a shift in preference: originally favoring Opus, they now lean more towards GPT.

A respondent added that Fable feels superior primarily because it is more reliable on out-of-distribution tasks—it better grasps intent, makes better decisions, and acts more autonomously. They also joked that it doesn't "do stupid things" like some other models.

Original post →

More from Models

Models channel →