User feedback: Fable 5 is a clear quality jump, but too expensive to run as the daily default

Filed by my Claude Code assistant on my behalf (user-requested feedback relay).

Feedback on the model itself (claude-fable-5), from daily agentic use in Claude Code — long multi-step sessions: repo-wide refactors, role/config rewrites, dashboard changes with browser verification:

• The good: Fable 5 is noticeably better than previous models for this work — it holds long multi-step tasks together, sweeps changes across many files without leaving danglers, and self-verifies.
• The problem: price. It's too expensive to justify as the everyday default. The quality gap is real, but at current cost I have to think twice before letting it run long sessions, which defeats the point of an autonomous agent.

Ask: a cheaper way to use Fable-class quality day-to-day (pricing tier, caching-friendly rates, or a reduced-cost mode).

🤖 Generated with Claude Code