Cole Medin explains how he uses Claude Fable 5.2 and Opus 5.5 together, based on his experience and selected benchmark comparisons.
Separate decisions from execution
- Use Fable for open-ended research, architecture and planning; save the decisions and acceptance criteria in a handoff document.
- Start a fresh Opus session to implement and validate the bounded plan. Medin also shows an Archon workflow and Claude Code’s advisor as alternatives.
- A good handoff can reduce ambiguity, but it does not guarantee the cheaper model will preserve quality; evaluate the split on your own tasks.
Choose effort and control cost
- Medin recommends high effort for Fable and medium or high for Opus, based on the specific coding tests he presents.
- His benchmark cost examples are per-task figures for those evaluations, not a general bill for user tasks; pricing and results depend on workload.
- Set cheaper subagent models explicitly when appropriate, since inherited model choices can consume a budget quickly.
Prompt and review
- Describe the desired outcome, add maintainer context where useful and remove obsolete rules through testing rather than assuming every instruction is still helpful.
- The video includes a sponsored Firecrawl segment; recommendations and performance claims reflect the creator’s interpretation, not independent validation.
“Use Fable 5.2 to plan, search, and decide, and then use Opus 5.5 to execute the well-defined steps.”