Grok 4.5 becomes the author’s daily default for agent work, with GPT-5.6 Sol for grounding

brandon_galang · x · 2026-07-25

The author says Grok 4.5 has become the default model in their daily workflow, especially for agent work.

They describe Grok 4.5 Medium Fast as the main model for light planning through general execution, while GPT-5.6 Sol is used when they need more careful grounding and Fable 5 for architecture/strategy review. They also say they liked Composer 2.5 Fast, but found it too weak to use beyond exploratory work.

Original post →

More from Models

Models channel →