Free guide · Sep 1, 2026
Fable 5.1's 'same price' fine print
What's really cheaper with Fable 5.1, and what isn't.
The short version
From the Anthropic docs page "What's new in Claude Fable 5.1", section Pricing:
Anthropic's own sentence: "Claude Fable 5.1 and Claude Mythos 5.1 are priced the same as Claude Fable 5, except for cache reads."
Fable 5 cache reads were \$1.00 / MTok (0.1 times base input, the rate every other Claude model uses). Fable 5.1 charges 0.025 times base input, so \$0.25. That is the whole price change. Batch API is \$5 in / \$25 out, half of base, unchanged.
For comparison, from the models overview table: Opus 5 is \$5 / \$25, Sonnet 5 is \$2 / \$10. All three have a 1M context window and 128K max output.
What's in the full guide
- 012. Cache math, worked
- 023. The effort dial, where the "3x cheaper" actually comes from
- 034. The seven behavior changes from Fable 5, and the fix for each
- 045. Three breaking API changes if you already call Fable 5
- 056. The Claude Code version you need
- 067. When to use Opus 5 instead
- 078. Benchmarks, vendor-reported
- 089. Fable 5.1 vs Mythos 5.1
- 0910. Other things in the docs worth knowing