Claude Fable 5.1 has a useful update for people building with AI: cheaper cache reads.
Anthropic announced it on September 1. Its listed cache-read price is $0.25 per million tokens, down 75% from Fable 5. That cut applies to cached input, not your entire bill.
Regular input and output prices remain $10 and $50 per million tokens. Anthropic estimates around 25% savings on typical token-billed workloads. Your result depends on how you use it.
Why this matters for a coding assistant:
A project can have a lot of repeated context. The same instructions. The same reference documents. The same background across several steps.
If your integration can reuse cached context, that part of the work can cost less.
A useful experiment: run the same small code-review task with your current model and Fable 5.1. Keep the instructions and files the same.
Record total cost, useful findings, mistakes and how much checking you had to do. Include cache usage in the comparison.
A lower token price is useful. A lower cost for work you can actually use is the better target.
This is a release summary and a suggested experiment, not a hands-on review.
Official details: https://www.anthropic.com/claude-fable-and-mythos-5-1 https://www.anthropic.com/claude/fable