PROMPTS
Fable 5 Is Here ⚡
Claude just dropped its most powerful model yet. Here is exactly what it is best for, when not to touch it, how to turn it on, and the 5 prompts to try on it.
Anthropic just released Claude Fable 5, and it is the most capable model they have ever shipped. But "most powerful" does not mean "use it for everything." The people who get the most out of Fable know exactly when to reach for it and when to leave it alone.
This guide is the whole picture: what it crushes, when it is the wrong tool, the two ways to turn it on, three things about how it works that nobody explains, and 5 copy-paste prompts to feel its power on day one.
Try It On The House
Fable 5 is the premium model, but it is free until June 22. That window is the perfect time to throw your hardest, biggest projects at it and see what it can do before you decide whether it earns a spot in your rotation.
What It's Actually Best For
Fable is built for the work that makes other models stumble: deep research, long documents, big multi-step projects, and complex builds. Anything where the task has a lot of moving parts and you need it to hold the whole thing in its head and not lose the thread halfway through.
The receipt: in testing, Stripe used Fable to run a code migration in a single day that would have taken a team of engineers over two months by hand. That is the kind of job it was made for. If a task feels too big to hand to AI, that is exactly the task to give Fable.
When NOT To Use It
Fable is the premium model. It costs about double to run and it burns through your usage faster. For a quick email, a simple rewrite, or a fast question, it is slower and pricier with no real payoff, you are paying for power you are not using.
Keep Sonnet or Opus for everyday work, and save Fable for the heavy lifting. The skill is not "always use the best model," it is "match the model to the size of the job." Reach for Fable when the task is genuinely hard or huge, not when you are firing off a one-liner.
How To Turn It On
In the Claude app: open the model picker at the top of your chat (where the model name shows) and choose Claude Fable 5. That chat now runs on Fable. Switch back the same way when you are done with the heavy task.
Building in Claude Code? Set your model to claude-fable-5. You can do it for one session with the /model command, or make it a project's default by adding "model": "claude-fable-5" to that project's .claude/settings.json. New to that file? My power-user settings guide walks through it.
Three Things To Know
With older models you had to decide how hard it should reason. Fable handles that itself. Its adaptive thinking is always on, so it decides how deeply to think based on the question, more for a hard problem, less for a simple one. There is nothing to toggle. You just ask, and it calibrates the effort on its own.
Fable can hold about 1 million tokens in mind at once. In plain terms, you can feed it huge documents, whole projects, or a year of notes in a single go and it keeps all of it in view while it works. This is why it is so good at the big stuff: it does not lose the early pages by the time it reaches the end.
On a small number of sensitive topics (under 5% of chats), Fable quietly hands the answer to Opus 4.8 instead. If you ever notice the response style shift on a touchy subject, that is what happened. It is a deliberate safety feature, not a bug, and for almost everything you do, you will never see it.
5 Prompts To Run On Fable First
These are built to show off what Fable does that smaller models cannot: hold a mountain of context, reason through complexity, and carry a long job end to end. Switch to Fable 5 first, then paste one in and fill in the brackets.
Fable can research wide and synthesize it into something you would actually pay a consultant for. Great while it is free.
This is where the 1 million token memory shines. Drop in a long contract, report, transcript, or PDF and let it actually read the whole thing.
Fable can carry a big multi-step build without losing the plot. Hand it the whole thing, not a tiny piece.
Dump everything, the brain dump, the voice-note transcripts, the scattered docs, and let Fable find the patterns you are too close to see.
The Stripe-migration energy, for whatever your version of a giant tangled mess is. Spreadsheets, a messy process, a system nobody understands anymore.
The Real Win
Fable 5 is not "a slightly better Claude." It is the model you reach for when a task is too big, too long, or too complex for anything else, and you leave it alone for the quick stuff. Learn that one habit, match the model to the job, and you get the best results without burning your usage. The free window until June 22 is your chance to feel the difference on your hardest project.
A prompt is only as strong as its context architecture. Always define: 1) Persona & Authority, 2) Precise Business Objective, 3) Constraints (what NOT to do), and 4) Structured Output Schema (tables, JSON, or step-by-step frameworks). Save your highest-performing prompt chains as reusable team SOPs.