Making Fable Cheaper Than Opus

Back to Devoured - July 14, 2026

DEVOURED

Making Fable Cheaper Than Opus

Cognition reduced costs for Devin by switching to the Fable 5 model, utilizing a proprietary fusion architecture that outperforms the previous Opus 4.8 despite higher per-token costs.

What: Joon Lee of Cognition announced that despite Fable 5's higher token price, structural changes to their agent's architecture resulted in lower overall costs and higher performance on FrontierCode 1.1 benchmarks.

Why it matters: This indicates that agentic cost management is increasingly about architectural efficiency rather than simply selecting the cheapest base model.

Deep dive

Architectural leverage: Cognition demonstrates that agent-level optimizations can negate higher unit costs of frontier models.
Performance benchmarks: The switch improved performance on FrontierCode 1.1 tasks.
Cost-benefit pivot: Companies are prioritizing total task-completion efficiency over raw cost-per-token metrics.

Original article

Making Fable Cheaper Than Opus

We replaced Opus 4.8 with Fable 5, and Devin’s bill went down. Fable 5 costs twice as much per token as Opus 4.8. But when we ran both models on FrontierCode 1.1 using our new Fusion architecture,...

Explore all 41 articles in today's digest