Using Fable 5 as an Advisor and Orchestrator

Using Fable 5 as an Advisor

In this scenario, Sonnet 5 acts as the executor and consults Fable 5 for recommendations.

Most of the tokens are billed at a lower executor rate.

On the SWE-bench Pro benchmark, the Sonnet 5 + Fable 5 combination as an Advisor achieves about 92% of Fable 5's quality, while the cost is approximately 63% of using a single Fable 5.

With this approach, Fable 5 is called infrequently (about once per task) to determine the strategy, while the bulk of the work is performed by Sonnet 5.

More details in the documentation.

Using Fable 5 as an Orchestrator

In this case, Fable 5 plans the execution of the task and distributes it among executors (Sonnet 5).

Most of the tokens are billed at a lower executor rate.

In BrowseComp tests, Claude Managed Agents were used with the Fable 5 orchestrator and Sonnet 5 subagents as executors. This configuration provides 96% of Fable 5's performance at 46% of its cost.

Resource-intensive research tasks are assigned to Sonnet 5.

More details in the example: cookbook.

P.S.

Claude Managed Agents supports both approaches.

With subagents, you can:

• delegate complex tasks to Fable 5 as an Advisor;
• assign task execution to Sonnet 5 executors.

Each subagent uses its own cache, so repeated calls do not require reprocessing the same context in full.