article --summary
The summary
The article looks at why Anthropic is steering most developers toward Opus 5.5 even though its newer Fable 5.2 model performs nearly as well on a coding benchmark and can finish certain tasks for roughly half the cost. Per-token prices favor Opus, but the reported total task cost for Fable came out lower, so token price alone does not decide which model is cheaper to run.
Anthropic also pitches Fable as an orchestrator that hands simpler jobs to Opus or Sonnet and steps in on harder problems. The author notes that this setup brings costs from extra model calls, context transfers, and failed attempts, and that thinking blocks from Fable cannot be read by Opus, which complicates handoffs. The piece suggests that developers may save money by switching to Fable earlier than Anthropic's capability-based rule advises.
Editorial note: This is a summary of an article first published by The New Stack. The full piece is on their site.