Claude Fable 5 (claude-fable-5) is the top of Anthropic's lineup, the most capable model it has made widely available. It is built for long-running agents and the hardest reasoning work, and it is priced accordingly. Here is what it costs, how it differs from Opus 4.8, and the cases where paying for it actually pays off.
The pricing, plainly
Fable 5 costs $10 per million input tokens and $50 per million output tokens. That is double Opus 4.8's $5/$25, and more than three times Sonnet 5's standard rate. There is no introductory discount, so the sticker price is the price.
It has a 1 million token context window and up to 128,000 output tokens on the standard Messages API, the same envelope as Opus 4.8. The size is not what you are paying extra for. The capability on hard tasks is.
Fable 5 vs Opus 4.8
They share the context window and the output cap, so the choice is about intelligence per dollar, not room to work.
- Opus 4.8 is Anthropic's model for complex agentic coding and enterprise work, at $5/$25. For most demanding tasks it is the sensible default.
- Fable 5 sits above it, aimed at the longest-horizon agentic runs and the most difficult reasoning. Anthropic positions it as next-generation intelligence for long-running agents, at the cost of higher price and slower responses.
Two behaviors are worth knowing before you send it a prompt. Thinking is always on, so Fable 5 reasons on every request and you cannot turn that off. And its safety classifiers can decline a request and hand it to a fallback model, which surprises people in week one. We wrote up exactly what triggers that and how to prompt around it in Claude Fable 5: what it can do, why it refuses, and the prompt fixes that stop the Opus fallback.
When the extra cost is worth it
At five times Sonnet 5's price, Fable 5 is not the model to reach for by reflex. It earns its rate on a narrow set of work: long autonomous agent runs where a single mistake compounds across many steps, and reasoning problems at the edge of what other models can do. For a listing description, a cold email, or a summary, a cheaper model gives you the same result for a fraction of the credits.
The trap is paying frontier rates for a task that never needed them. The way to avoid it is to compare the output on your own prompt rather than assume the priciest model is the right one.
How to run your prompt against Fable 5 in FixMyPrompt
FixMyPrompt lets you take your prompt and run it against Fable 5 directly, so you can see whether the top-tier model actually produces a better answer for your task, or whether Opus 4.8 or Sonnet 5 gets you there for less.
- Open your prompt in FixMyPrompt.
- Pick Claude Fable 5 from the model list in the test run.
- Run it, then run the same prompt against a cheaper model and compare the two outputs.
It is pay as you go, so you spend a few credits to answer the question instead of committing to the most expensive model on faith. On the hard, high-stakes prompts where Fable 5 pulls ahead, you will see it. On the everyday ones where it does not, you will save the difference.
Run your prompt against Claude Fable 5 in FixMyPrompt. No signup needed to try it.