What each plan gets
Fable has been a paid extra on Pro since July, and on standard Team seats too, billed to usage credits on top of the plan. The launch offer puts every Fable model back in those plans from 9 to 16 November. Fable's cap is half the weekly limit, taken from the same limit the other models use, and unused room doesn't carry over. Pro's one free reset has the same deadline. After it, Fable on Pro goes back to usage credits.
Max and Ultra plans and premium Team seats already had Fable in the plan, under the same cap. Their offer is free limit resets, usable through 8 March 2027 Pacific time. How many depends on how long you've paid. Max gives five times Pro's weekly usage, and Ultra ten times.
1
free reset on Max, Ultra or a premium seat paid for under six months
Source: Claude's help page on the Fable 5.2 launch offers (read November 11, 2026)
2
free resets after six to twelve months paid
Source: Claude's help page on the Fable 5.2 launch offers (read November 11, 2026)
3
free resets after more than a year paid
Source: Claude's help page on the Fable 5.2 launch offers (read November 11, 2026)
Who qualifies for a reset
To get a reset, you must have been paying by 10:00 am PT on 9 November, launch day. Claude's help page sets no such cutoff for Pro's included week, so a Pro plan started later still gets Fable at no extra cost, without the reset.
- On a Team plan the cutoff applies to the organization's billing, and the seat holder had to join before it too.
- A plan still in a free trial at the cutoff gets its resets once it converts to paid. Self-serve Team trials count as paid.
- Time paid counts from your first payment, or from the app store's first charge if you subscribed there. Cancelling and coming back doesn't restart it, and on a Team plan the organization's first payment counts, so every premium seat gets the same number.
- Cancel or downgrade before using a reset and it's gone. Resets have no cash value and can't be moved to another account.
- Free plans get no offer at all, and most Enterprise plans don't either.
Last verified on 11 November 2026 at 03:25 UTC against Claude's help pages under Sources.
How to claim it
There's nothing to sign up for. Fable 5.2 doesn't become your default model, though, so pick it each time you want it.
- On Pro, open Settings > Usage and switch Usage credits off if you don't want to pay for Fable. Fable then stops at its cap and the other models carry on. Left on, Fable keeps going and the extra is billed to your usage credits.
- In Claude on the web, the desktop app or the mobile app, choose Fable 5.2 in the model picker. Claude Cowork needs the latest Claude Desktop for it.
- Or, in Claude Code, run
claude updateto get version 2.1.321 or later, then/model fable. It starts on Medium effort, and/effort highraises it. - For a reset, press Reset for free under Limit resets in Settings > Usage, then press it again to confirm. A used reset can't be undone.
Steps from Claude's help pages and Claude Code's model docs. Mindket hasn't used a reset itself.
What a reset restores
In Anthropic's own example, an account has used 80% of its weekly limit, 50 points on Fable and 30 on other models, and a reset takes it back to 30% used. The weekly refill still comes at its usual day and time. On an account with a five-hour limit, a reset also refills that one completely, other models' usage included. Verified Ultra accounts have no five-hour limit, so for them it touches only the week.
You don't have to wait for a limit, though the reset button also shows on the message you get when you reach one. Claude Code in a terminal or an IDE has none, but the limits belong to the account, so a reset made in the Claude app restores your Claude Code allowance too.
Sample builds from X, with their time and cost
On a live stream, Riley Brown gave Claude one prompt for a short film. Fable 5.2 wrote code that built and animated a room in Blender, 13 camera shots of it, then rented 10 GPUs on RunPod, where the 4K render took 52 minutes. Music and sound came from ElevenLabs.
Mr Tactical gave Fable 5.2 on High and Opus 5.5 the same prompt for a glass-pool scene in 3D. Fable took 90 minutes to Opus 5.5's 68 and cost four times as much. Mr Tactical rated its sunset and night scenes well above Opus 5.5's.
Marcel sent one pelican prompt to two different tools, so the times aren't a controlled comparison. Fable 5.2 on High, in Claude Code, took about 25 minutes to draw a pelican riding a bicycle through five animated seasons in SVG. Google's Gemini 4 Hydrogen, tried on Arena before its launch, took about 5.
Felix Rieseberg, who works at Anthropic, gave Fable 5.2 one-shot prompts and no feedback for 8 hours, and it made a first-person shooter with its own models, sounds and textures. No other model got the same brief, so it shows what Fable can make, not how it compares.
| Sample | What it made | Model and effort | Time | Cost given |
|---|---|---|---|---|
| Riley Brown | A 4K film, built and rendered in Blender | Fable 5.2 | 52 minutes to render | $5.75 in total |
| Mr Tactical | A glass-pool scene in 3D | Fable 5.2, High | 1 hour 30 minutes | $57.87 |
| Mr Tactical | The same scene | Opus 5.5 | 68 minutes | $14.39 |
| Marcel | An animated pelican in SVG | Fable 5.2, High, in Claude Code | About 25 minutes | About $25 |
| Marcel | The same SVG | Gemini 4 Hydrogen, on Arena | About 5 minutes | Not given |
| Felix Rieseberg | A first-person shooter | Fable 5.2 | 8 hours | About $550 at API prices |
| Anthropic | An operating system that boots on a laptop | Fable 5.2, over 1,000 subagents | 8 days | Not given |
Its prose still draws complaints
A course Fable 5.2 built for Ethan Mollick on W. H. Auden was "accurate and joyless" until several rounds of feedback on taste. Mollick's review says its prose kept slipping into "Claudespeak", the review's name for the habit. Mollick put under an hour into a research paper it wrote and called the result "not publishable as it is".
Dimitris Papailiopoulos, a researcher at Microsoft, posted that Fable 5.2 "writes worse than Opus 5.5" and has slid back into "Slopus", a jab that merges slop and Opus.
Ahead of Opus 5.5 on most of Anthropic's tests
On FrontierCode 2.0, which asks whether open-source maintainers would merge a model's code, Anthropic's announcement has Fable 5.2 at 49.1% at Xhigh effort, nearly four points above Opus 5.5's best. Google's Gemini 4 Carbon wins AutomationBench, a test of business workflows, 53.9% to 51.4%, and Opus 5.5 keeps a 15-point Elo lead on AA-Briefcase, a ranking of knowledge work.
Artificial Analysis, which benchmarks models independently, puts Fable 5.2 first on its index. Its post on X has it level with Opus 5.5 on knowledge work and terminal coding: on Terminal-Bench 4.0 it scores 60%, behind Kimi K3.1 Pro's 67%.
| Test | Fable 5.2 | Fable 5.1 | Opus 5.5 | GPT-6 Astra | Gemini 4 Carbon |
|---|---|---|---|---|---|
| FrontierCode 2.0, code maintainers would merge | 49.1% | 32.4% | 45.5% | 42.8% | 41.1% |
| FrontierSWE v2, software engineering | 69.9% | 56.3% | 62.3% | 65.5% | 61.2% |
| OSWorld 2.1, using a computer | 89.3% | 83.5% | 87.2% | 79.5% | 81.5% |
| Humanity's Last Exam, with tools | 74.2% | 65.6% | 67.7% | 57.2% | Not given |
| Chartography, reading charts | 73.4% | 44.8% | 64.4% | 71.0% | 72.5% |
| AutomationBench, business workflows | 51.4% | 31.4% | 42.5% | 41.4% | 53.9% |
| RoboDojo Sim, robot control | 64.9% | 27.6% | 39.4% | 29.8% | Not given |
Source: Artificial Analysis' post on X about Fable 5.2 (read November 10, 2026)
A task took 17.5 minutes on High and 44 on Max
Fable 5.2 writes its answers fast. ClaudeDevs, Anthropic's developer account, announced it as up to 30% faster than Fable 5.1. Artificial Analysis measured 103 output tokens a second at Max, and @IndependentEco averaged more than 100 over a 30-minute session and found it faster than Opus 5.5. An optional fast mode speeds output up by as much as 2.5 times again, at double the price per token.
That speed didn't make tasks short. At Max, the first word of an answer took about 235 seconds to arrive, and the average task in Artificial Analysis' test took 44.4 minutes, 71st of the 72 models timed. It also wrote 275 million output tokens across the test, the second most of 73 models.
On Artificial Analysis' index, High scored 62, three points under Max, in 17.5 minutes a task. The same test puts Max's cost per task at about 2.6 times Opus 5.5's.
Anthropic's own FrontierCode chart, a different test, shows Max costing more for less. Fable at High matched Opus 5.5's best score there, 45.3% to 45.5%, for roughly half the cost per task. Max scored three and a half points below Xhigh, at the highest cost on the chart.
Yuchen Jin of Databricks reposted the chart with one line: "Don't use Fable 5.2 Max man".
| Effort | Intelligence Index | Minutes per task | Output tokens per second |
|---|---|---|---|
| Max | 65 | 44.4 | 103 |
| Xhigh | 62 | 26.1 | 103 |
| High | 62 | 17.5 | 93 |
| Medium | 56 | 8.1 | 88 |
| Low | 54 | 6.5 | 92 |
Screenshots of X's own embed, taken on 10 November 2026, with times in UTC. Mindket hasn't timed Fable 5.2 itself. Select a shot to open the whole post, which the embed cuts at Show more.
What a month of Fable on Ultra would cost on the API
Anthropic doesn't say how much of the weekly limit a Fable task takes. Its help page says only that Fable models use limits faster than its other models, with no multiple.
Theo, who runs T3 Chat, measured one Claude Ultra account's limit from 1.1 million requests sent through a proxy between 11 August and 9 November. A month of Fable's cap, spent entirely on Fable 5.2, would cost about 15 times the plan's price at Anthropic's API list rates. That's what the same work would cost on the API, and it's money saved only if you'd otherwise pay those rates. On the API, Fable 5.2 costs a fifth less per token than Fable 5.1 and twice as much as Opus 5.5.
@kimmonismus calls Fable 5.2 the best model they've used, one that fixes bugs Opus 5.5 missed. Their complaint is the limits: it eats through them fast, the five-hour one most, even on Medium effort.
| Model | Input, per million tokens | Output, per million tokens |
|---|---|---|
| Fable 5.2 | $8 | $40 |
| Fable 5.2, fast mode | $16 | $80 |
| Fable 5.1 | $10 | $50 |
| Opus 5.5 | $4 | $20 |
| Sonnet 5.5 | $2 | $10 |
Try Fable on High with one job you can check
On Pro, spend the free week on one job whose result you can check, such as a bug that has resisted fixes or a migration you've put off. Switch Usage credits off first. Going by the effort tests, we'd run it on High and move to Xhigh only if High's result fails that check. Leave everyday coding and writing on Opus 5.5, as Anthropic advises, since it stretches the same limit further. Save the reset for the day Fable stops you mid-job with the weekly refill still days away.
On Max or Ultra, unused resets last until 8 March 2027. Keep them for a heavy week, before a launch or a deadline, and give Fable the work Opus 5.5 got wrong first.
Frequently asked questions
Why can't I see Fable 5.2 in the model picker?
On a Team or Enterprise plan, an owner decides which models members see. Enterprise plans keep Fable 5.2 off until an admin turns it on. If your role there can already use a Fable model, Fable 5.2 joins it seven days after launch, unless an admin opts out first. A Free account has no Fable models at all. In Claude Cowork, Fable needs the latest version of Claude Desktop.
Do the offers cover Fable 5.1 and Fable 5 too?
Yes. Usage on Fable 5.1 or Fable 5 counts toward Fable's half of the weekly limit, and a reset gives it back along with Fable 5.2's.
Does a reset refund usage credits I've already spent?
No. Charges already made stay as they are, and a reset restores only the limits the plan itself comes with.
Do Enterprise plans get either offer?
Most don't. Self-serve and usage-based Enterprise plans get neither offer, and neither do standard seats on legacy seat-based Enterprise plans. Premium seats on those legacy plans get one to three resets. HIPAA-ready Enterprise plans get none.
Sources
- Claude Help Center: Claude Fable 5.2 launch offers https://support.claude.com/en/articles/17506494-claude-fable-5-2-launch-offers
- Claude Help Center: Claude Fable models on your plan https://support.claude.com/en/articles/15424964-claude-fable-models-on-your-plan
- Claude Help Center: What is a limit reset? https://support.claude.com/en/articles/17007452-what-is-a-limit-reset
- Claude Help Center: Usage credits https://support.claude.com/en/articles/12429409
- Claude Code docs: Model configuration https://code.claude.com/docs/en/model-config
- Claude: Plans and pricing https://claude.com/pricing
- Anthropic: Introducing Claude Fable 5.2 and Claude Mythos 5.2 https://www.anthropic.com/claude-fable-and-mythos-5-2
- Riley Brown on X: a Blender film from one prompt https://x.com/rileybrown/status/2119902109709537708
- Mr Tactical on X: Fable 5.2 and Opus 5.5 on one glass-pool scene https://x.com/MrTacticalX/status/2119955647542542439
- Marcel on X: the pelican SVG test against Gemini 4 Hydrogen https://x.com/marcthecreatorr/status/2120151349924938083
- Felix Rieseberg on X: a shooter from one-shot prompts https://x.com/felixrieseberg/status/2119925633564463609
- Ethan Mollick on X: Portals, 3D worlds made in code https://x.com/emollick/status/2119918307457577163
- Alex Albert on X: a da Vinci notebook made interactive https://x.com/alexalbert__/status/2119916954337349737
- Dom on X: app screens from Fable 5.2 https://x.com/dominikmartn/status/2120154832694858224
- Ethan Mollick, One Useful Thing: Fable 5.2 and Jagged AGI https://www.oneusefulthing.org/p/fable-52-and-jagged-agi
- Dimitris Papailiopoulos on X: Fable 5.2's writing https://x.com/DimitrisPapail/status/2119954897365795062
- Artificial Analysis: Claude Fable 5.2 models, intelligence, performance and price https://artificialanalysis.ai/models/releases/claude-fable-5-2
- ClaudeDevs on X: Fable 5.2 in the Claude Platform and Claude Code https://x.com/ClaudeDevs/status/2119868698613956992
- @IndependentEco on X: a 30-minute speed test https://x.com/IndependentEco/status/2119880979414466647
- Yuchen Jin on X: Anthropic's FrontierCode chart by effort https://x.com/Yuchenj_UW/status/2119867942603264279
- Artificial Analysis' post on X about Fable 5.2 https://x.com/ArtificialAnlys/status/2119878995386982508
- Epoch AI's post on X: Fable 5.2 on the Epoch Capabilities Index https://x.com/EpochAIResearch/status/2119903415363113088
- ARC Prize's post on X: Fable 5.2 on ARC-AGI https://x.com/arcprize/status/2119871080525373922
- Theo on X: what a month of Claude Ultra holds at API prices https://x.com/theo/status/2119960608816988596
- @kimmonismus on X: Fable 5.2 and the five-hour limit https://x.com/kimmonismus/status/2120076998383227133
- Claude API docs: Pricing https://platform.claude.com/docs/en/about-claude/pricing