Skip to content
Nubent

/Anthropic · Nov 9, 2026

Claude Fable 5.2 arrives on Amazon Bedrock

Anthropic released Claude Fable 5.2, built for demanding reasoning and long-running agentic work, with a 1M-token context window and a lower price per token than version 5.1.

What changes

Fable 5.2 brings a 1M-token context window, up to 128k output tokens and always-on adaptive reasoning. It costs USD 8 per million input tokens and USD 40 per million output tokens, versus USD 10 and USD 50 for Fable 5.1. It's available on the Claude API and on Amazon Bedrock, among other platforms. It also always enforces the prefix check: if your code edits the system prompt, the tools or an earlier message and then resends a reasoning block from a later turn, the request returns a 400 error.

How to implement it

  1. Check whether your code modifies the system prompt, the tools or earlier messages between turns: with 5.2 that returns an error.

  2. Run your evaluation set on real cases, comparing 5.1 and 5.2 on quality, latency and cost.

  3. Change the model ID in configuration, not in code, so you can roll back quickly.

  4. Measure actual output tokens before assuming the savings: always-on reasoning consumes tokens too.

Nubent's take

The price cut is appealing, but the critical part is the prefix check: it breaks integrations that rewrite history to save context. That's the first thing we'd review before migrating.

Next step

Should we implement it in your account?

We check whether it fits, how much work it takes and what risks it carries. 30 minutes with an engineer.