Claude Fable 5 for developers: key API changes and pricing shifts

Anthropic has launched its first generally available Mythos-class model, Claude Fable 5, which brings significant shifts in pricing and API behavior. Released on June 9, 2026, the model is priced at $10 per million input tokens and $50 per million output tokens—exactly double the rates of Claude Opus 4.8. While the cost increase is straightforward, the real challenge lies in the API’s breaking changes, which could silently disrupt existing integrations if not addressed proactively.

As operators of model intelligence at TokenMix, where we monitor pricing and API behavior across over 300 models, we’ve verified every detail in this guide against Anthropic’s official documentation, migration guide, and pricing page as of June 10, 2026.

Key updates at a glance

Before diving into the technical changes, here’s a quick rundown of what developers need to know:

• Pricing structure: All rates are exactly 2× those of Claude Opus 4.8. For example, cache reads cost $1 per million tokens, while 5-minute cache writes are $12.50 per million tokens.
• Model identifiers: The model is accessible as claude-fable-5 via the Claude API, anthropic.claude-fable-5 on AWS Bedrock, and anthropic/claude-fable-5 on OpenRouter.
• Context and output limits: The model supports 1 million tokens of context and 128,000 tokens of output, with no surcharges for long contexts.
• Zero-data retention (ZDR) limitation: Accounts without 30-day data retention enabled won’t have access to the model at all.

Three breaking changes demand immediate attention:

1. Adaptive thinking is permanently enabled, and the `thinking: {