Fable vs Sol in real work: Fable orchestrates, Sol executes
The writer subscribes to both and GPT Pro and uses Claude Code (Fable) and Codex (Sol) daily for real work. Sol needs more hand-holding than Fable — it makes more mistakes and struggles to grasp the bigger picture. For a big task like writing a full client proposal, Sol isn't trustworthy yet.
In terms of how close the output gets to what's needed, Fable reaches roughly 80% of the way, while Sol lands around 60-65%. Still, Sol is an improvement over the earlier Opus model. Sol's strength is narrow, well-defined tasks.
Based on this, the best is having Fable orchestrate the overall work while Sol executes the narrow, focused pieces.
Key points
- Both (Fable via Claude Code) and GPT Pro (Sol via Codex) used side by side in real work
- Sol makes more mistakes and needs closer supervision than Fable
- Sol isn't yet trusted for big tasks like a full client proposal
- Rough feel: Fable ~80%, Sol ~60-65% (though Sol beats the earlier Opus model)
- Recommended : Fable orchestrates, Sol executes narrow tasks