Grok 4.6 surprisingly good, while Fable feels heavily quantized

teortaxesTex · x · 2026-08-30

User comparison of Grok 4.6 and Fable: Grok 4.6 performs surprisingly well at times, while Fable recently feels heavily quantized, exhibiting hallucinations (e.g., claiming China has 4x Japan's population) and failing to maintain context in a long game design discussion. The user speculates this may be to drive hype for a new model.

Original post →

More from Models

Models channel →