Claude Fable 5 is a Claude model. Of everything in the release, five upgrades change how writing, research and code work day to day, and the rest you can read about later. Below are those five, the way I prompt Fable differently, and ten prompts you can run today.
What Claude Fable 5 is
Fable 5 is a Claude model, and it is the version this guide and its ten prompts are tested against. The details that move fastest, which plans include it and what it costs through the API, live in the guide so I can update them without rewriting this page.
If you are new to Claude altogether, start with how to use Claude and come back. This page assumes you know what a project and an artifact are, and that you have run a few real jobs through a chat.
The five upgrades that matter
A model release ships a long changelog. Five items change how you work, and the guide's chapter The 5 Upgrades That Actually Matter is the filter. Below are the five dimensions I check on every Claude release, with what the guide reports for Fable 5 on each. Every item carries a marker until the number behind it is confirmed against the release notes, because I would rather show you the gap than fill it with a guess.
1. How much it can hold at once. In practice this decides whether you hand Fable a chapter or the whole book, one function or the whole repo.
2. How well it keeps to the rules you set. Format instructions given once at the top of a long piece should still hold on page nine.
3. Code and tool use. For anyone using Claude Code, this is the upgrade that shows up first, as fewer retries and fewer half-finished edits.
4. Thinking before answering. Whether the model can take longer on a hard problem when you ask, and how you switch that on.
5. Speed and price. A model that is cheaper and faster changes which jobs you automate and which you keep doing by hand.
If one of those five is the reason you are here, the guide has the specifics. If none of them is, your current Claude setup probably serves you fine, and the prompts below still work.
How to prompt Fable 5 differently
The habits below are general good practice, and they matter more on a newer model because it does more with less scaffolding. These are the six changes I made to my own prompts.
Say the whole job once. Put the goal, the audience, the material and the format in the first message. Drip-feeding context made sense when models lost the thread; on a model built for long work it only costs you turns.
Ask for the plan before the output. On anything longer than a page, "outline first, wait for my yes, then write" catches a wrong direction in ten seconds instead of ten minutes.
Retire the tricks that were workarounds. Prefilled answers, elaborate tag structures and "think step by step" nudges were patches for older models. Test each one before you keep it.
Tell it what to do when something is missing. "If a fact is missing, write MISSING in brackets and keep going" stops the model from inventing a number to fill a hole. I run my own articles this way.
Set the format once, then hold it to the format. If the output drifts, quote the rule back rather than restating the whole prompt.
Ask for the objection. "What is the strongest argument against this draft?" gets more use out of the model than a fourth polish pass.
Ten prompts you can run today
The guide carries the full text of each. Here is what each one does, by job, so you can pick the one you need.
Writing
- The voice-matched newsletter. Paste two past issues and the topic. The prompt studies sentence length, openers and sign-off, then drafts the third issue in that voice.
- The landing page from a fact sheet. Feed it product facts and the audience. It returns headline options, the sections in order, and the objection each section answers.
- The plain-language rewrite. Paste any draft. It strips the generic phrasing, varies sentence length, and adds one specific detail per paragraph, then shows the before and after.
- The transcript to post. Paste a call or video transcript. It pulls the two or three ideas worth a post and drafts each one under a set word limit.
Research
- The evidence-backed summary. Paste a long PDF or report. Every claim in the summary comes with the sentence it was drawn from, so you can check it.
- The comparison table. Give it two or more vendor pages or docs. It builds a table on the criteria you name and flags where a claim is missing.
- The plan review. Paste a plan. It argues against it as a skeptical investor, lists the assumptions, and ranks which ones would hurt the most if wrong.
Code
- The error explainer. Paste the stack trace and the function. It names the cause in plain words, proposes the fix, and says what to check afterwards.
- The review against a checklist. Paste a diff and your team's rules. It reviews only against those rules and cites the line for each finding.
- The script from a spec. Describe the job in plain English. It writes the script, the tests, and the one command to run it.
Each of these works in a chat and in Claude Code. Once one of them becomes a habit, the Claude Skills mini course shows how to turn it into a skill Claude loads on its own.
What is inside the free Claude Fable 5 guide
- Claude Fable 5 Prompt Engineering Guide
- 10 Ready-to-Use Fable 5 Prompts
- The 5 Upgrades That Actually Matter
- Lifetime access and updates
Leave your email and the PDF arrives in your inbox. When the model changes in a way that affects the prompts, the guide changes with it, at no extra charge.
For ready-made prompts by task beyond these ten, the Claude prompts in the library are sorted by job.
