Opus 5.5 vs Fable 5.1

The default, and the step up

Anthropic's documentation frames this choice directly: Opus 5.5 is the default for most work, and Fable 5.1 is the step up for the hardest long-running agentic and research tasks — or for work where Opus 5.5 at a higher effort still falls short. Its launch post goes further and places Opus 5.5 at Fable 5.1's level on most work. That makes the burden of proof clear: Fable 5.1 has to earn its price on a task Opus 5.5 has already been tried against.

What the step costs

The headline gap is wide — the table shows Fable 5.1's input and output rates well above Opus 5.5's. The cache-read gap is not: both models discount reads deeply, and their cache-read rates sit close together. So on a workload dominated by cache reads, the premium for Fable 5.1 is concentrated in output and fresh input tokens, and in the fact that long Fable turns on hard tasks tend to produce a lot of both.

Constraints that decide it before cost does

Fable 5.1 requires data retention, so an organisation on zero data retention can't call it without express authorisation. Neither model offers Priority Tier. Both reject forced tool choice and disabled thinking, and both bind thinking blocks to the conversation — and on the Claude API, Fable 5.1 can read Opus 5.5's thinking blocks, while no model but Mythos 5.1 reads Fable 5.1's. That asymmetry matters for routing: escalating a conversation from Opus 5.5 to Fable 5.1 keeps its reasoning; falling back the other way does not. (The documentation states the first half for the Claude API only, so don't rely on it on partner platforms.)

Choosing

Run the task on Opus 5.5 at high or xhigh first. If it still falls short in a way that matters, try Fable 5.1 on the same inputs and compare cost per completed task, not per token.

Related

See claude-fable-5-1, claude-opus-5-5 and Fable models and zero data retention.