Claude Fable 5 is expensive, so Anthropic now recommends using it mainly as a planner while handing execution off to smaller models. The developer team outlines two strategies. In the “Advisor” pattern, Sonnet 5 runs as the executor and only calls Fable 5 when it needs guidance. On SWE-bench Pro, this combo reaches about 92 percent of Fable 5’s solo performance at 63 percent of the cost. In the second pattern, Fable 5 acts as a planner that delegates tasks to Sonnet 5 worker agents. On BrowseComp, this delivers 96 percent of Fable 5’s performance at 46 percent of the cost. Both patterns run through Claude Managed Agents, with each sub-agent using its own cache to avoid duplicate context costs.
This approach signals growing price pressure on Anthropic as Chinese open-source models undercut Western pricing and new GPT-5.6 Sol offers lower per-token costs. The shift suggests companies will prioritise cost efficiency over raw model capability for routine tasks.
* Fable 5 is called roughly once per task in the Advisor pattern
* Each sub-agent uses its own cache to avoid duplicate context costs
* Official documentation contains further details on the implementation
