Anthropic ships Claude Fable 5.1 with 75% cheaper cache reads

Anthropic put Claude Fable 5.1 on general release September 1, leaving base API prices alone but slashing the cost of a cache hit by three quarters, and it reports sizable jumps over Fable 5 on agentic research, coding and business-workflow benchmarks.

Key points

• Base rates are unchanged from Fable 5 at $10 per million input tokens and $50 per million output, while a cached read drops to $0.25 per million from $1; batch jobs run at half price, $5 in and $25 out.
• Anthropic expects the cheaper cache to make Fable 5.1 about a quarter cheaper to run than Fable 5 on a typical workload, and up to about 45% cheaper on the most agent-heavy work.
• On Terminal-Bench-Science 0.1 the new model posts 52.6% against Fable 5's 24.7%, and on AutomationBench it posts 31.4% against 17.1%.
• The model ID on Anthropic's API is claude-fable-5-1, and the same model is offered through AWS, Google Cloud and Microsoft Azure; Anthropic's announcement adds that Claude Code runs it at High effort by default, with Medium the default on Claude.ai and Cowork.
• Mythos 5.1 is the same model with some safeguards relaxed for cyber-defense and life-science work, and only approved organizations can use it; Anthropic's announcement says access is currently limited to a set of US organizations while it works with the US government on widening that.

Why it matters

If you pay for Claude Code or your own agents by the token, a 75% cut on cache reads hits your bill directly, and it matters most in agent-heavy work, where the same context is read back again and again. Fable 5.1 is not a default in Claude Code, though: its model docs say no plan or provider starts on a Fable model, so you opt in with /model fable, and once selected it runs at High effort unless you change it. Anthropic's model page points most workloads to Opus 5 first and keeps Fable 5.1 for harder reasoning and long-running agent work.

This is an AI-written summary of the reporting credited above and the other sources linked in the text, read and edited by Nicholas before publishing. The facts and any quote belong to those sources; the wording is ours. Read the original.