How to prompt Claude Fable 5.1

When a task is genuinely hard, it is tempting to reach straight for the biggest model. Claude Fable 5.1 is Anthropic’s most capable tier, and also its slowest and most expensive. Anthropic’s own advice is clear: start somewhere else, and move up to Fable when you have evidence you need it. This guide shows you how to tell, and how to brief Fable well once you get there.

Is this guide for you? If you are new to AI, you probably do not need Fable yet. Many weak answers come from vague requests, and a clear brief with our AIM method helps every modelModelA specific version of an AI, such as Claude Opus 5.5 or GPT-6 Astra. Different models trade speed, cost and ability..

✦

1

The idea in 30 seconds

Good at

The hardest reasoning and long, many-step jobs where an early mistake is costly.

Watch the 30-second explainer

The idea in 30 seconds0:00 / 0:26

1. Good at The hardest reasoning and long, many-step jobs where an early mistake is costly.
2. Pick it when You tried Opus 5.5, worked harder on the brief, and it still fell short.
3. 3 habits Ask for a plan and updates. Ask for plain words. Show one example of correct quoting.
4. Check before you trust Fluent, confident answers make errors harder to spot. Check every source.
5. Climb the ladder: brief well, try Opus, then Fable.

Key takeaways

If you remember nothing else, remember these.

• Start with Claude Opus 5.5. Move to Fable 5.1 for demanding reasoning and long agent work, or when Opus still falls short.
• Fable writes fewer progress updates, denser prose and less formatting than earlier models. Ask for what you want instead of carrying over old rules.
• In summaries it may reuse source wording without quotation marks. For academic work, show it one example of correct quoting.
• Ask for targeted edits, not whole rewrites, and set clear limits on what it may change.

Is Fable 5.1 the right model for you?

Anthropic released Claude Fable 5.1 on 1 September 2026. Its official description is a model for demanding reasoning and long-running agent work. Compared with the previous Fable, Anthropic highlights stronger long-running coding, multistep research, and work on documents, spreadsheets and slides.

Here is where it sits in the lineup. Cost is relative, based on the price developers pay; app plans are priced differently.

Anthropic’s models overview says that if you are unsure, start with Opus 5.5. It points to Fable for demanding reasoning and long agent work, or when Opus at a higher setting still falls short. That gives you a simple ladder:

1. Brief well first. Write a proper AIM brief with a precise action verb.
2. Run Opus 5.5. Our Opus 5.5 guide covers how to brief it.
3. Give Opus more thinking time if your app or tool lets you, and check the result against your own criteria.
4. Then try Fable 5.1. Compare the two on the same task and the same checks.

The picker below opens on a hard factory problem where a strong model already fell short. Switch the first answer to “Hard, multi-step” and watch the recommendation drop back to Opus.

Interactive

Which Claude model fits your task?

Tap an example, or answer the four questions yourself. The suggestion on the right updates as you go.

Start from an example

Label a question bank by Bloom levelScreen 400 abstractsRoute SME WhatsApp messagesDraft a CLO-aligned rubricRun a literature-review agentFind why a Penang factory’s yield droppedMatch a month of supplier invoices to payments

1How hard or new is the task?

RoutineSome judgementHard, multi-stepI tried a strong model; not enough

2How many items, how often?

One-offA batch of dozensHundreds, or ongoing

3How fast must each answer come?

I can waitWithin a minuteInstantly

4What shape is the work?

A single answerA few steps, or one step in a bigger agentA long-running agent

Recommended

Claude Fable 5.1

SpeedSlowerPrice$10 / $50

Anthropic positions Fable 5.1 for demanding reasoning and long-horizon agent work, or for when Opus 5.5 at higher effort still falls short. It is the slowest and most expensive tier, so confirm Opus cannot do the job first.

Starting effort (how hard it thinks)

lowmediumhighxhighmax

default

High is Fable's default and the recommended start. Step down to medium or low only where your own checks show quality holds.

Read the guide

• Fable 5.1$10 / $50
• Opus 5.5$4 / $20
• Sonnet 5.5$2 / $10
• Haiku 5.5from $0.10 / $0.50

Price in US$ per million tokens (input / output)

In the Claude apps you pick the model; effort is a setting developers choose in the API. Either way, the official advice is to test on your own tasks rather than copy settings. This picker is our rule of thumb, not an official decision tree.

Effort, in plain words

Before answering, Fable always does some private thinkingThinking (reasoning)Private working-out the AI does before it replies, a bit like rough notes before a final answer., and effortEffort / reasoning effortA setting that tells the AI how hard to think before answering. Higher effort is slower and costs more, but helps with difficult work. sets how much. Picture a quick glance at the accounts versus a full audit. Fable starts high, close to the full audit, because that is what it is for.

Drag the slider, or tap a level, to trade speed for depth.

Speed and savingsThinking depth

• Recommended start
• For the most demanding work

lowmediumhigh★ Defaultxhighmax

Thinking depth3/5

Speed and savings3/5

Relative scale, for illustration.

low

Often competitive with Opus and Sonnet on cost per task. Calls search tools less often, so nudge it to search.

mediumRecommended start

Roughly matches the previous Fable model at lower cost. Use it for routine work once quality holds.

highDefaultRecommended start

The default. Start here, then test the other levels on your own tasks.

xhighFor the most demanding work

For the most demanding agent and coding work. It may think longer before a long deliverable.

maxFor the most demanding work

Deepest and slowest. Leave plenty of room for output.

Two practical points come from Anthropic’s guide:

• “Expensive” depends on the job. Anthropic says Fable at its lowest setting is often competitive with Opus and Sonnet on cost per task while scoring higher. So the biggest model is not always the dearest choice.
• The very top settings can mean long waits. At the highest levels, Fable may draft a long report in its head and then write it out again. For a long report or rewrite, its default is usually right.

In the Claude app you mostly choose the model, and whether Fable appears in your menu depends on your plan. If your app or tool also shows a thinking or effort option, leave it at the default unless you have a reason to change it.

In the app: habits that get better answers

These habits come from Anthropic’s Fable 5.1 prompting page, turned into everyday requests you can type in a chat.

1. Ask for a plan, updates and a recap. Fable goes quiet on long tasks. Ask for a one-line plan before it starts, brief updates as it goes, and a short recap at the end that makes sense on its own.
2. Ask for plain words. Its sentences can run long, with few paragraph breaks. Anthropic’s short fix is the request “Please remove all mannered prose.” You can add “use short paragraphs and plain words”.
3. Say when a list or table helps. Earlier models overused bullets and bold, so many old prompts ban them. Fable already leans the other way. Remove old bans on MarkdownMarkdownSimple symbols (#, *, -) the AI uses for headings and bullets. formatting and say when you want a table.
4. Show one example of correct quoting. When summarising, Fable is more likely than before to reuse source wording without quotation marks. Anthropic’s fix is one complete example of a correct answer, with a sentence on why it is correct.
5. Ask for targeted edits. Fable may rewrite a whole document for a small change. Ask it to change only the lines that need it.
6. Set the scope. It may fix nearby things or stretch the task. Say what is in scope, and ask it to list extra ideas at the end instead of doing them.

Here is how those habits change a research request:

Old habitsFable 5.1 style

You are a research methodologist helping a PhD candidate write a literature review.

The 40 attached papers study AI adoption among Malaysian SMEs. I will cite your output in my thesis, so quoting must be clean. Example of correct quoting: "Owners cited cost as the main barrier (P07), while P12 found skills mattered more." Own words, source IDs, quotation marks only for exact phrases.

Synthesise the findings into 4 to 6 themes. For each, give two or three plain paragraphs and the paper IDs that support it. Put any exact wording in quotation marks with its paper ID. Use a table only for the theme overview. Finish with a short recap and a list of papers you could not classify.

What you typically get back

Themes you can check against the papers, every claim tied to a source ID, exact wording marked as quotation, and a recap that stands on its own.

Ready-to-copy prompts for local work

Fable earns its price on long, multi-step work where an early mistake is expensive. Each prompt uses AIM plus a precise action verb.

You are a senior auditor who helps small finance teams find errors in their books.

<ledger_export>
(paste the trial balance and the general ledger for March, with staff and supplier names removed)
</ledger_export>
The trial balance is out by RM1,240. Last month's balance agreed. We have already checked the bank reconciliation, and it agrees.

Diagnose the most likely causes. Before you start, state your plan in one line. For each possible cause, say what in the ledger would confirm or rule it out, and check it. Change nothing; only report. Finish with a ranked list of causes and the single check I should do first.

You are a research assistant who reports uncertainty honestly.

I have uploaded 30 full-text papers on why Malaysian polytechnic graduates leave technical jobs, plus my draft framework. Use only these files and say when the evidence is thin.

Compare the evidence for at least three competing explanations, including any that contradict my framework. State your plan in one line first and give a brief update after each explanation. End with a recap table: explanation | supporting papers | contradicting papers | strength of evidence.

You are an OBE curriculum reviewer preparing a programme for MQA self-review.

I have uploaded 12 course files, each listing CLOs mapped to our 7 PLOs, and a file listing the PLOs. Do not invent PLO numbers or accreditation rules I have not given you.

Audit every CLO-PLO mapping. Flag CLOs whose verb does not match the PLO's level, and PLOs covered by fewer than two courses. Suggest targeted edits only to the flagged lines, never rewrite whole files, and list every suggestion in a table: course | old line | new line | reason.

The trial-balance card shows why Fable fits diagnosis: it tests each possible cause instead of jumping to the first one. For the full agent pattern, compare these with the agent briefing guide.

When Fable works on its own

If your organisation runs Fable as an AI agentAI agentAn AI that does a multi-step job for you (planning, opening files, using tools and checking its work) instead of only replying once. on long, unattended jobs, two more habits from Anthropic’s guide matter.

It may stop early or ask permission it already has. On long autonomous work it sometimes describes the next step instead of doing it, or asks “Shall I apply this?” for something you already requested. If no one is watching in real time, say so. Tell it to carry on with steps that can be undone, and to stop only for destructive actions or real changes of scope.

Ask for a short explanation, not its full reasoning. Prompts that ask Fable to write out all its thinking may be declined. Ask for a short explanation of the answer, or a summary of the steps it took.

Practise on a real prompt. Tap each highlighted habit to see what the Fable 5.1 guidance says, then apply the fix.

Interactive

Modernise an old prompt

Tuned for Claude Fable 5.1

This lecturer's prompt is full of habits from older models. Tap each numbered line to see why the official advice says to change it, then tap Apply this fix.

Old-style prompt

0/8

1. 1Default roleYou are a helpful assistant.
2. 2Shouted ruleCRITICAL!!! You MUST NEVER go over ONE page!!!
3. 3"Think step by step" lineThink step by step and write out all your reasoning before the answer.
4. 4Rule with no reasonUse simple English.
5. 5Vague requestHelp me with a rubric for my marketing presentation.
6. 6Old anti-formatting ruleNever use bullet points, headers or bold. Plain paragraphs only.
7. 7Prefilled answerAssistant: Sure! Here is your rubric:
8. 8No definition of done(nothing says when the job is finished)

Habit 1 of 8

Default role

Why change it

Being helpful is already the default, so this line adds no expertise, no standard and no audience.

What the official guidance says

Anthropic recommends giving Claude a role; even one sentence focuses its tone and behaviour. Make the role specific.

The fix

You are an assessment designer who writes fair, clear rubrics for first-year diploma students in Malaysia.

Apply this fixNext habit

Modernised prompt

Fix allResetCopy prompt

0 of 8 habits fixed

Apply a fix to start building the modern prompt.

Effort (a setting, not text you type): high is the default for Fable 5.1. Step down to medium or low for routine work once you have seen quality hold; keep xhigh and max for the most demanding work.

For developers: technical detailsOptional: skip this if you only use the chat app

For developers using the APIAPIA way for software to talk to an AI directly, without the chat app. Developers use it to build AI into their own systems.. The model ID is claude-fable-5-1; it is also available on Amazon Bedrock, Google Cloud and Microsoft Foundry.

• Specs and price. A 1-million-token context windowContext windowHow much text an AI can keep in mind at once: your instructions, the files you share and the conversation so far., up to 128K output tokensTokenThe small chunks of text an AI reads and writes: roughly half to three-quarters of an English word, depending on the model. Pricing and limits are counted in tokens., and adaptive thinking that is always on. US$10 / US$50 per million input / output tokens (Opus 5.5: US$4 / US$20). Batch requests are 50% cheaper.
• Claude Mythos 5.1 is the same model with different safeguards, available only to organisations verified through Anthropic’s programmes.
• Effort. The default is high, and it is the recommended start. At medium, Fable 5.1 roughly matches Fable 5 at lower cost; at low, it is often competitive with Opus and Sonnet models on cost per task while scoring higher. Reserve xhigh/max for measured gains. At low it calls search and retrieval tools less often.
• Long outputs at xhigh/max can be drafted in thinking and then written again; run them at high or leave generous max_tokens, because it caps thinking and reply together.
• Progress updates arrive as progress-update thinking blocks that are empty under the default display; set display: "updates" (beta) to receive them, and ask in the system promptSystem promptStanding instructions set up before the conversation starts, such as "always answer in Bahasa Melayu". They apply to every reply. for updates at predictable points.
• Keep the conversation history append-only. Editing earlier turns, or changing the system prompt or tools mid-session, invalidates thinking blocks.
• Forced tool use returns an error on Fable 5.1.
• Safety classifiers can return a refusal; handle stop_reason: "refusal" in your code.
• Autonomy. Anthropic publishes two system-prompt additions for long asynchronous work: one telling the model it is operating autonomously and should proceed with reversible steps, and one defining the user’s request as the scope.

{
  "model": "claude-fable-5-1",
  "max_tokens": 64000,
  "output_config": { "effort": "high" }
}

See all model guides on the models page.

Check your understanding

Check yourself

Fable 5.1 in five questions

1/5

Pick an answer, then tap Check answer. You can also use the arrow keys to move between answers.

Question 1 of 5: A colleague wants to use Fable for every task "to be safe". What does Anthropic suggest instead?

AStart with Opus 5.5 and move to Fable when Opus still falls shortBUse Fable for everything; it is always bestCUse Haiku for everything to save money

Check answer

Sources

6

1. 1Anthropic: Prompting Claude Fable 5.1 (opens in a new tab)platform.claude.com
2. 2Anthropic: Claude Fable 5.1 overview (opens in a new tab)platform.claude.com
3. 3Anthropic: Models overview (opens in a new tab)platform.claude.com
4. 4Anthropic: Effort (opens in a new tab)platform.claude.com
5. 5Anthropic: Prompting best practices (opens in a new tab)platform.claude.com
6. 6Anthropic: Prompt engineering overview (opens in a new tab)platform.claude.com

Nice work!

Guide complete

You made it to the end. Take the quiz above if you have not yet, for bonus XP.

• 0 XP
• 0 day streak