Model Delegation Reshapes AI Agent Cost Structures

The high cost of AI agents is a major pain point for businesses. Anthropic launched its flagship Fable 5 model in June 2026, pricing it at $10 per million input tokens and $50 per million output tokens—twice the price of Opus 4.8, released in May. To address the problem, technology companies are exploring architectures that delegate complex tasks to less expensive models.

OpenRouter benchmarks conducted in June 2026 found that a Fusion delegation architecture pairing Fable 5 as the lead model with assistant models reduced API usage costs by 54%. In complex tasks, the configuration also outperformed and outscored a setup pairing Opus 4.8 with assistant models. The results showed that delegated workflows can deliver better performance at a lower cost.