Claude Fable 5.1 API & Pricing Guide

01

API access and model identifier

Anthropic lists Claude Fable 5.1 on the Claude API as claude-fable-5-1. Anthropic also lists availability through Amazon Web Services, Google Cloud and Microsoft Foundry.

Provider-specific identifiers can differ, so verify the current identifier in the provider documentation instead of assuming the native Claude API string works everywhere.

02

Context and output limits

Anthropic documents a 1 million-token context window and up to 128K output tokens for Fable 5.1. Adaptive thinking is always on, with effort controlling reasoning depth.

Treat the large context ceiling as capacity, not a target. Sending unnecessary context can raise latency and spend without improving task success.

03

Claude Fable 5.1 API pricing

Anthropic currently lists $10 per million input tokens and $50 per million output tokens. Five-minute cache writes cost $12.50/M, one-hour cache writes cost $20/M, and cache reads cost $0.25/M. Batch API input and output receive a 50% discount.

The cache-read rate is 75% below Fable 5’s $1/M rate. Anthropic estimates about 25% lower cost for typical Fable workloads and up to roughly 45% for highly agentic workloads; those are vendor estimates, not guaranteed savings.

04

Plan cost around cache reuse

Measure uncached input, cache writes, cache reads, output length and completed-task success. Context-heavy agents that repeatedly reuse repositories or tool definitions can benefit more from $0.25/M cache reads than short single-turn tasks.

Compare total cost per successful task against Fable 5 and cheaper Claude tiers rather than comparing token prices in isolation.

05

Migration checks before production

Anthropic documents breaking behavior around forced tool use and thinking-block compatibility. Forced tool selection can error, earlier models cannot read Fable 5.1 thinking blocks, and editing earlier turns invalidates thinking blocks.

1. Replace the model with the provider-specific Fable 5.1 identifier.
2. Confirm account-specific limits and provider settings.
3. Measure cache-hit behavior before estimating savings.
4. Regression-test tool choice, thinking blocks and retries.
5. Run a representative quality set at the effort levels you plan to use.

06

Data retention can affect deployment

Anthropic’s current model documentation says Fable 5.1 requires 30-day data retention and is not generally available with zero data retention unless Anthropic expressly authorizes the organization.

For regulated or privacy-sensitive workloads, verify contractual and provider-specific retention terms before routing production data to the model.

07

Bottom line

Fable 5.1’s API economics are most attractive for long, context-heavy agents because base token rates stay unchanged while cached context is substantially cheaper to reread.

For the broader model overview, see the Claude Fable 5.1 guide; for migration tradeoffs, compare Fable 5.1 with Fable 5 before shifting production traffic.