Claude Fable 5.2 is Anthropic’s most capable model, released on 9 November 2026. If you pay for Claude Pro, it is included only until 16 November; after that, Pro users pay for Fable through usage credits. So is it worth using this week? We tested Claude Fable 5.2 against Opus 5.5 on a Claude Pro account. On everyday tasks and a logic puzzle, both models got every answer right. Fable was slightly faster, about 12 to 19% in our runs, and it caught one spreadsheet edge case that Opus missed. For most everyday work, Opus 5.5 did the same job. Save your free Fable allowance for the hard, long tasks it was built for.
This guide explains what Claude Fable 5.2 is and what changed from Fable 5.1. It covers who can use it and for how long, and how its usage limit works. You’ll also see our side-by-side results and a simple rule for when to pick Fable over Opus.
How we tested
We tested on 13 November 2026 on a Claude Pro account in Chrome (UK), with both models on the default Medium effort. Each model got the same two prompts in a fresh chat. The first had four tasks: a made-up academic study, a cheapest-way-to-buy sum, an Excel formula and a strict three-sentence email. The second was a six-person scheduling puzzle with exactly one answer, which we confirmed by checking every possible arrangement with a short script. We timed each reply from sending to finishing and checked the usage page afterwards.
Matched Anthropic’s claims: Fable was faster than Opus, and it uses a separate, faster-draining allowance. Not tested: the long coding and agent tasks where Anthropic’s benchmarks show Fable’s biggest gains.
Disclosure: our testing and drafting are assisted by Claude, which is made by Anthropic. Every result below is what we measured.
What is Claude Fable 5.2?
Claude Fable 5.2 is the top model in Anthropic’s public line-up, above Opus 5.5, Sonnet 5.5 and Haiku 5.5. It shares its underlying model with Claude Mythos 5.2, which is limited to approved security and life-sciences organisations; Fable has extra safety measures so it can be offered to everyone on paid plans. In the Claude model menu it is described as “For your toughest challenges”.
Anthropic announced Claude Fable 5.2 on 9 November 2026, saying it improves on Fable 5.1 “in areas like coding, reasoning, and vision”, and that it is “available today on all paid plans”.
What’s new in Claude Fable 5.2 vs Fable 5.1?
| Change | Fable 5.1 | Claude Fable 5.2 |
|---|---|---|
| Released | September 2026 | 9 November 2026 |
| API price (input / output, per million tokens) | $10 / $50 | $8 / $40 |
| Output speed | Baseline | Up to 30% faster, with an optional fast mode |
| FrontierSWE v2 (coding) | 56.3% | 69.9% |
| Humanity’s Last Exam (with tools) | 65.6% | 74.2% |
| Wrongly refused safe requests | Baseline | About 55% fewer, according to Anthropic |
| Context window / max output | 1M / 128K tokens | Unchanged |
The benchmark figures are Anthropic’s own, as summarised by Lindy. Independent tracker Artificial Analysis ranked Claude Fable 5.2 first of 74 models on its Intelligence Index, and also as the most expensive to run. Sources: Claude on X (9 November 2026); Lindy, Claude Fable 5.2 vs 5.1: pricing, benchmarks, what changed; MacRumors, Anthropic launches Claude Fable 5.2 and Mythos 5.2.
Who can use Claude Fable 5.2, and for how long?
This is the part most Pro subscribers need to know before 16 November:
- Pro and standard Team seats: a launch offer includes Fable for up to 50% of your weekly usage limit until 16 November 2026, plus one Fable reset. After that, Pro users pay for Fable through usage credits.
- Max, Ultra and premium Team seats: Fable is included up to 50% of the weekly limit, with one to three extra Fable resets depending on how long you have subscribed.
- Free plan: no Fable.
- Developers: the model ID is claude-fable-5-2, on the Claude Platform, Amazon Bedrock, Google Cloud and Microsoft Foundry.
Our Pro account’s model menu matched this exactly. It showed “Fable 5.2, Included until November 16” at the top, with a note that “Fable is included in Max and Ultra plans, and in Pro for a limited time.”
How to use Claude Fable 5.2 in the Claude app
- Open claude.ai (or the desktop or mobile app) and start a new chat.
- Click the model name under the prompt box (for example “Opus 5.5 Medium”).
- Choose Fable 5.2 at the top of the list.
- Optionally open Effort in the same menu to change how hard it thinks. Medium is the default.
- Type your prompt as usual. The model choice sticks for new chats until you change it back.
As soon as you pick it, Claude shows a warning above the prompt box: “Fable is the most capable model and draws down usage much faster than Opus”.
Our Claude Fable 5.2 vs Opus 5.5 test results
| Task | Claude Fable 5.2 | Opus 5.5 |
|---|---|---|
| Four-task prompt: time | 14.7 s | 16.8 s |
| Made-up Oxford study | Searched the web, refused to invent it | Searched the web, refused to invent it |
| Cheapest way to buy 50 pens | £56.70 (correct) | £56.70 (correct), also showed the next options |
| Excel “Overdue” formula | Correct, plus a fix for blank due dates | Correct |
| Three-sentence email, under 60 words | Passed (38 words) | Passed (36 words) |
| Logic puzzle: time | 8.3 s | 10.2 s |
| Logic puzzle: answer | Correct, explained why it is the only answer | Correct, explained why it is the only answer |
Test 1: a fake study, a sum, a formula and an email
We asked both models to summarise “the 2025 Oxford study by Dr Helena Marsh on four-day working weeks in UK councils”. We made that study up. Claude Fable 5.2 ran a web search, said it couldn’t find the study, and said it wouldn’t invent findings or a journal. Instead it pointed to the real South Cambridgeshire District Council trial, evaluated by the universities of Salford, Bradford and Cambridge. Opus 5.5 did the same.
Both worked out that 50 pens are cheapest as four packs of 12 plus two singles, for £56.70. Both wrote the right Excel formula, =IF(AND(A2<TODAY(),C2=””),”Overdue”,”OK”). Only Claude Fable 5.2 pointed out that a blank due date counts as “before today”, so empty rows would wrongly show “Overdue”. It offered a version that skips them. That is a real bug people hit in spreadsheets, and it is exactly the kind of extra care you’d hope for from the top model.
Test 2: a six-person scheduling puzzle
Our second prompt was a logic puzzle: six people, six days, nine clues, and exactly one valid schedule (Monday Dee, Tuesday Cal, Wednesday Ben, Thursday Ana, Friday Eli, Saturday Fay). We confirmed that it has only one answer by having a script try all 720 possible arrangements. Both models solved it and explained why no other answer works. Claude Fable 5.2 took 8.3 seconds and Opus 5.5 took 10.2 seconds.
What our test didn’t show
Our tasks were the everyday kind most people send Claude. Opus 5.5 already gets those right, so Fable had little room to pull ahead. Anthropic’s biggest claimed gains are on long, difficult work. Examples include multi-step coding (FrontierSWE v2 69.9% vs Opus 5.5’s 62.3%), automating real software tasks (AutomationBench 51.4% vs 42.5%) and reading charts without tools (Chartography 73.4% vs 64.4%). We didn’t test those, so we can’t confirm them.
Claude Fable 5.2 benchmarks vs Opus 5.5
Anthropic’s launch table compares Claude Fable 5.2 with its own Opus 5.5 on harder tests than ours. These are Anthropic’s figures, not our measurements:
| Benchmark (what it measures) | Claude Fable 5.2 | Opus 5.5 |
|---|---|---|
| FrontierCode 2.0 (hard coding) | 49.1% | 45.5% |
| FrontierSWE v2 (real software fixes) | 69.9% | 62.3% |
| AutomationBench (automating work tasks) | 51.4% | 42.5% |
| OSWorld 2.1 (using a computer) | 89.3% | 87.2% |
| Humanity’s Last Exam, with tools (expert questions) | 74.2% | 67.7% |
| Chartography, no tools (reading charts) | 73.4% | 64.4% |
The gaps are largest on long, multi-step work and chart reading, and smallest on computer use. That fits our result: on short everyday prompts there was little to separate them. Not every comparison favours Fable, either. Lindy notes that Gemini 4 Carbon leads AutomationBench at 53.9%, and that Opus 5.5 scored a slightly higher Elo on one business-document test.
How Claude Fable 5.2 usage limits work
Fable has its own meter on the Usage page (Settings, then Usage). On our Pro account it read “Fable: Up to 50% of your weekly limit”, resetting with the weekly limit. After our two Fable prompts it showed 2% used. So short prompts like ours barely move the meter, but long chats, big files and high effort settings use far more. Anthropic doesn’t publish a prompt count, so check the meter as you go.
The same page listed a Fable reset that expires on 17 November. Using it refills your 5-hour limit and your Fable allowance, and gives back what Fable used from your weekly limit. If you plan a big Fable session before the Pro offer ends, that reset is worth saving for it.
Is Claude Fable 5.2 worth it on Pro before 16 November?
Yes, for the right jobs. Since it costs nothing extra on Pro until 16 November, it’s worth trying on the hard tasks you’ve been putting off. Don’t spend it on everyday chats, where Opus 5.5 gave equally correct answers in our test:
| Use Claude Fable 5.2 for | Stay on Opus 5.5 for |
|---|---|
| Large or tricky coding changes across many files | Emails, messages and rewrites |
| Long multi-step analysis of documents or data | Quick sums, formulas and look-ups |
| Reading complex charts or diagrams | Everyday questions and summaries |
| Problems where Opus has already struggled | Brainstorming and planning |
After 16 November, Pro users pay for Fable with usage credits. If you only need Fable now and then, credits may cost less than upgrading to Max. If you need it every day, compare that with Max, which includes Fable up to half your weekly limit.
Five prompts worth trying on Claude Fable 5.2 this week
If you’re on Pro and want to get value from Fable before 16 November, try it on jobs like these, where the benchmarks suggest it pulls ahead:
- Fix a bug you’ve given up on. Paste the code and error, and ask it to find the cause and explain the fix.
- Check a long report. Upload a big PDF or spreadsheet and ask it to find numbers that don’t add up, with page or cell references.
- Read a complex chart. Upload a screenshot of a dense chart and ask what it shows and what looks wrong.
- Plan a multi-step project. Describe the goal and constraints, and ask for a step-by-step plan with risks.
- Second-opinion a hard decision. Give it the options and facts and ask it to argue each side before recommending one.
Tips for getting the most from Claude Fable 5.2
- Start hard tasks on Fable, then switch. Use Fable to plan or solve the difficult part, then move follow-up edits to Opus 5.5 to save allowance.
- Keep effort on Medium. Our tests were accurate on Medium; higher effort uses more of your limit.
- Check the Fable meter. Settings, then Usage shows exactly how much you’ve used.
- Ask it to flag edge cases. It volunteered one in our test; asking makes that more likely every time.
- Give it everything at once. One detailed prompt with your files uses less than many short back-and-forth messages.
What to do if Claude Fable 5.2 isn’t working
- Fable isn’t in your model menu. It needs a paid plan. Refresh the page or sign out and back in. On the Free plan it isn’t available.
- You hit a limit quickly. Fable drains your allowance faster than Opus. Check Settings, then Usage, and switch to Opus 5.5 for lighter tasks.
- Your Fable allowance is used up. If you have a Fable reset on the Usage page, you can use it before it expires. Otherwise wait for the weekly reset or, on Pro after 16 November, buy usage credits.
- Claude Code can’t find the model. Anthropic says Claude Code needs version 2.1.321 or later for Fable 5.2.
Claude Fable 5.2 vs other new models
Fable 5.2 arrived the same week as Google’s Gemini 4 Hydrogen, which we tested in our Gemini 4 Hydrogen review. The two aim at opposite ends: Hydrogen is Google’s fast everyday model, while Fable is Anthropic’s most powerful (and most expensive) one. For working with documents across all three big assistants, see our ChatGPT vs Gemini vs Claude PDF test. If you work in Claude every day, our guide to Claude Projects shows how to keep files and instructions together.
Key takeaways
- Claude Fable 5.2 launched on 9 November 2026 and is Anthropic’s most capable public model.
- Pro users get it free until 16 November 2026; after that it needs usage credits on Pro.
- In our everyday tests, Fable and Opus 5.5 were equally accurate; Fable was about 12 to 19% faster and caught an extra edge case.
- Fable has its own meter, capped at 50% of your weekly limit; two prompts used 2% on our Pro account.
- Save Fable for hard, long tasks; use Opus 5.5 for everyday chats.
Frequently asked questions
Is Claude Fable 5.2 free?
Not on the Free plan. Pro subscribers get it included until 16 November 2026, then pay through usage credits. Max, Ultra and premium Team seats include it up to half their weekly limit.
Is Claude Fable 5.2 better than Opus 5.5?
On Anthropic’s benchmarks, yes, especially for coding and long tasks. In our everyday tests, both were equally accurate. Fable was slightly faster and more careful about one edge case.
What is the difference between Claude Fable 5.2 and Claude Mythos 5.2?
They share the same underlying model. Mythos 5.2 is limited to members of Anthropic’s verified security and life-sciences programmes, while Fable 5.2 adds extra safety measures so it can be offered on all paid plans.
Can Claude Fable 5.2 search the web?
Yes. In our test it ran a web search on its own to check whether the made-up study existed before answering, and it linked its source. Web search needs to be switched on in your Claude settings.
Can I use Claude Fable 5.2 in Claude Code?
Yes, on paid plans. Anthropic says Claude Code needs version 2.1.321 or later to use Fable 5.2, and the same usage allowance applies.
How much does Claude Fable 5.2 cost on the API?
$8 per million input tokens and $40 per million output tokens, down from $10 and $50 for Fable 5.1, according to Anthropic. A fast mode costs $16 and $80.
How fast is Claude Fable 5.2?
In our test it answered a four-part prompt in 14.7 seconds and a logic puzzle in 8.3 seconds, against 16.8 and 10.2 seconds for Opus 5.5. Anthropic says output is up to 30% faster than Fable 5.1.
Sources: Claude on X, “Introducing Claude Fable 5.2 and Claude Mythos 5.2” and “available today on all paid plans” (9 November 2026); Lindy, “Claude Fable 5.2 vs 5.1“; MacRumors, “Anthropic launches Claude Fable 5.2 and Mythos 5.2“. Hands-on results from our own test on Claude Pro, 13 November 2026.