Building on the Claude Platform: Claude Fable 5 and model orchestration patterns

The Claude lineup is wider than it has ever been: Fable 5 for frontier reasoning and long-horizon agentic work, Opus 4.8 for everyday complex tasks, Sonnet 5 as the workhorse default, and Haiku for speed and scale. This session gives developers a framework for thinking through how to use Fable as an advisor and when to delegate out to other models. 

Join Brad Abrams and Jeremy Hadfield, members of Anthropic’s Technical Staff, for a look at how to build an eval suite from your own tasks and use it to decide which work moves to Fable 5. From there, we walk through the advisor strategy, an orchestration pattern where a smaller, cheaper model does the work and Fable 5 sets the strategy, often matching frontier-level results at a fraction of the token cost. You'll see how the pattern runs on the Claude Platform along with eval results on intelligence and cost.

Who should attend: Heads of AI, platform engineering and architecture leaders who own model strategy and internal evals, and developers building agents and orchestration on the API.