---
title: Anthropic Releases Claude Fable 5.2 at a 20% Lower Price, With Up to Three Free Resets for Subscribers
description: Anthropic prices Fable 5.2 at $8 and $40 per million tokens and gives Max and Ultra subscribers one to three free resets by length of subscription.
author: Darie Nani (Editor-in-Chief)
date: 2026-11-10T13:04:40.551Z
updated: 2026-11-10T13:09:36.719Z
canonical: https://www.sovereignmagazine.com/article/claude-fable-5-2-price-free-resets
image: https://cdn.nanimediahouse.com/claude-fable-5-2-anthropic-launch-card.webp
categories: Artificial Intelligence
content_type: News
region: United States
publication: Sovereign Magazine
about:
  - type: Organization
    name: Anthropic
    description: American artificial intelligence company that develops the Claude models.
---

Anthropic released Claude Fable 5.2 on 9 November at $8 per million input tokens and $40 per million output tokens, 20% less than the previous version, and is giving paying subscribers up to three free resets of their Fable usage.

Fable is the most expensive model Anthropic sells to the general public, and the one it describes as its most capable. After the cut it still costs twice as much as Opus 5.5, which Anthropic calls "our strongest daily driver" and recommends as the starting point for most coding and knowledge work. Anthropic says Fable is best for "the most ambitious, open-ended technical work".

Fable 5.2 produces output up to 30% faster than Fable 5.1, according to Anthropic, and its safety filters blocked about 55% fewer requests in the company's early testing. With the lower price and the free usage, the model is cheaper and easier to try than Fable 5.1 was. It is the same model as Mythos 5.2, which Anthropic supplies without those filters to vetted organizations working in cybersecurity and biomedical research. In July we reported on [the launch of Fable 5](https://www.sovereignmagazine.com/article/anthropic-fable-5-fumbled-launch), which was withdrawn three days after it went on sale and returned on 1 July.

## Early Fable 5.2 Demos Include a Shooter Made From One Prompt for About $550

![A first-person view of a container port at sunset in the rain, with a rifle sight aimed at a flying drone, from a game made with Claude Fable 5.2](https://cdn.nanimediahouse.com/fable-5-2-shooter-demo.webp)
*A level from the first-person shooter that Anthropic engineer Felix Rieseberg made with Claude Fable 5.2 from a single prompt. — Photo: Felix Rieseberg*

Felix Rieseberg, an engineer at Anthropic, posted an example on the day of the release that has been viewed more than a million times on X. He asked for a first-person shooter in a single prompt and gave the model no feedback while it worked. "Everything in this game is from scratch: models, sounds, textures," he wrote. The run took eight hours and used about 445 million input tokens and 2.7 million output tokens, which he put at about $550 at API prices.

Rieseberg also published the prompt, which was three sentences long. In it he asked for "something that looks visually like it can keep up with AAA", the games industry's term for big-budget titles, and for the scope to be kept to one small level. The result is [playable in a browser](https://claude.ai/artifact/BGY1SztrAYcJQHSsnHHGmp). The level is a container port in the rain under a low sun, where the player has to cut three relays while fighting flying drones and a four-legged robot. A compass, an ammunition count and a health reading are shown on screen.

Riley Brown, a founder and YouTuber, said he gave the model one prompt during a live stream. It wrote code that built a room in the 3D software Blender, animated a short film there across 13 camera shots, rented 10 graphics processors from the cloud service RunPod and rendered the film in 4K in 52 minutes, according to his post on X. "Total cost: $5.75," he wrote.

Lex Toumbourou, who blogs about coding agents and had early access to the model, gave it his nephew's drawings of characters for a fighting game and a prompt before going to bed. "When I woke up, the game was waiting for me," he wrote. The model had generated about 175 images and 140 audio clips with outside image and sound services he let it use, and had written a program that played all 12 stages of the game and checked screenshots for problems. Toumbourou made further changes with the model after his nephews had played it.

Anthropic says one of its engineers had Fable 5.2 coordinate more than 1,000 subagents for eight days to produce a prototype operating system that boots on a laptop. The company also says the model designed three working electronic devices from a one-paragraph prompt and wrote out the proof of Fermat's Last Theorem in a form a computer can check, in 30 hours. Those are the company's own accounts.

Artificial Analysis, which tests AI models independently, ranked Fable 5.2 first on its measure of interactive 3D scenes on 10 November. Human panelists compare pairs of scenes without being told which model made them. On their votes Fable 5.2 was rated 1,271, against 1,202 for Opus 5.5 and 1,168 for OpenAI's GPT-6 Astra. Peter Gostev, who works on AI capability at the ranking site Arena, wrote that Opus "made nice looking games but gaming mechanics were sloppy", while Fable "can one shot games that are actually fun".

I asked Fable 5.2 to redesign Sovereign's website in one go, setting it to work for an hour before I went to bed, and went through the result over coffee the next morning. The run used about 12% of my weekly allowance on my Ultra plan. The new design went live with very minimal changes and I hope readers like it as you are looking at it.

## Fable 5.2 Costs Twice as Much as Opus 5.5 After the Price Cut

> "I often still prefer Opus 5.5 for everyday work. Fable's got a big brain and thinks a little slower."
> — Felix Rieseberg, Anthropic

Rieseberg wrote that in a post on the day of the release, in which he said Fable "beats Opus on big, long tasks" such as rewriting an entire codebase or producing a first draft of a large piece of work. Dianne Penn, Anthropic's head of product management for research, told CNBC that Fable 5.2 is intended for "power users or power use cases".

Fable 5.1 was priced at $10 and $50 on the same measure. Opus 5.5 costs $4 and $20, half the new Fable price. An optional fast mode, which generates text up to 2.5 times faster, doubles Fable's price to $16 and $80.

On Anthropic's own benchmark table, Fable 5.2 leads Opus 5.5, OpenAI's GPT-6 Astra and Google's Gemini 4 Carbon on most tests. On FrontierCode 2.0, which measures whether a model's code changes would be accepted by the maintainers of open-source projects, it scores 49.1% against 45.5% for Opus 5.5, 42.8% for GPT-6 Astra and 41.1% for Gemini 4 Carbon, which Anthropic ran at a lower effort setting. Gemini 4 Carbon, [which passed Opus 5.5 on one independent index earlier this month](https://www.sovereignmagazine.com/article/google-gemini-4-carbon-ranks-first-cyber-defenders), leads on AutomationBench, a test of business workflows, with 53.9% against 51.4% for Fable 5.2.

Anthropic argues that Fable does more for the money on hard tasks. At its high effort setting, the company says, Fable 5.2 scores 45.3% on FrontierCode, about level with the best Opus 5.5 result, at roughly half the cost per task.

During training Anthropic "took steps to limit advances in the specific research capabilities that could most directly accelerate AI development", the company says. On tasks such as tuning the low-level software that runs on AI chips and improving the architecture of language models, Fable 5.2 therefore performs at about the level of Opus 5.5. Chief executive Dario Amodei [called in September for the development of frontier AI to be paced](https://www.sovereignmagazine.com/article/dario-amodei-pace-the-frontier-embedded-evaluators). "We called for pacing, not pausing," Penn told CNBC.

## Fable Resets Are Counted From a Subscriber's First Paid Bill

Subscribers on the Max and Ultra plans, and holders of premium seats on Team plans and older seat-based Enterprise plans, receive one reset if they have paid for Claude for less than six months, two for six to 12 months and three for more than 12 months. Anthropic counts from the first paid bill on the account, "even if you paused or switched plans since", according to [its help page](https://support.claude.com/en/articles/17506494-claude-fable-5-2-launch-offers). A subscriber who paid for the $20 Pro plan for a year and has since moved to the $200 Ultra plan therefore receives three, the same as someone who has paid for Ultra throughout. On Team and Enterprise plans Anthropic counts from the organization's first bill, so every premium seat in a company gets the same number.

Those resets can be used at any time until 8 March 2027, and a subscriber who cancels or downgrades before using one loses it. Accounts needed paid billing in place before the launch, at 10am Pacific time on 9 November.

Pro subscribers and holders of standard Team seats normally pay for Fable through separately bought usage credits. For one week, until 16 November, they can use it inside their plans for up to half of their weekly limit, and they get one reset that expires the same day. Free users get nothing.

After a reset, the part of the weekly allowance spent on Fable is restored. Usage of other models stays where it was, and the limits themselves do not rise. In Anthropic's example, a subscriber who has used 80% of the weekly limit, 50 points of it on Fable, is back at 30%. On plans with a five-hour limit, that window is refilled as well. Fable use stays capped at half of the weekly limit, [as it was when the $200 plan became Ultra](https://www.sovereignmagazine.com/article/claude-ultra-plan-5-hour-limit). Anthropic gives no reason for the offers beyond marking the launch.

## Fable Uses Up Plan Limits Faster Than Other Claude Models

Anthropic warns on its help page that "Fable models use your usage limits faster than other Claude models". Gostev said one of the dozens of prompts he ran on Fable 5.2 cost $2,618. "It's a great model, a LOT of things got better but not every single thing, it does make sense to use Opus still," he wrote.

Theo Browne, a developer and YouTuber who runs the chat app T3 Chat, said Fable 5.2 made a game for him in 76 minutes at a cost of about $42 in tokens. He called it "not significantly better than Opus 5.5 (esp. for the price)", and later wrote that Fable 5.2 "is making a lot of stupid mistakes I would not have expected from Opus".

Pro subscribers who want to keep using Fable after 16 November will pay for it through usage credits. Max and Ultra subscribers keep it inside their plans at up to half of the weekly limit.

Find out more at [anthropic.com](https://www.anthropic.com/claude/fable).

## FAQ

**Q: Is Claude Fable or Opus better?**
Fable 5.2 scores higher than Opus 5.5 on most of the tests Anthropic published and on Artificial Analysis's blind ranking of 3D scenes, but it costs twice as much per token and uses plan limits faster. Anthropic recommends Opus 5.5 for most coding and knowledge work and Fable for long, open-ended technical tasks.

**Q: What is the difference between Fable and Mythos?**
They are the same underlying model. Fable 5.2 has safeguards that block high-risk cybersecurity work and most research biology tasks, and it is available on paid Claude plans and through the API. Mythos 5.2 comes without those safeguards and is supplied only to organizations Anthropic has vetted through its cyber and life sciences verification programs.

**Q: Why does Claude keep switching from Fable to Opus?**
When Fable's safety classifiers block a request, an older model answers instead: Opus 4.8 for cybersecurity and Opus 5 for biology. A small set of AI development topics also falls back to Opus 5. Anthropic says Fable 5.2 was blocked about 55% less often than Fable 5.1 in early testing.