Anthropic released Claude Fable 5.2 and its restricted sibling Mythos 5.2 on Monday, unveiling AI models the company says can turn a handful of instructions into working software, design physical objects, and tackle research problems that previously required teams of specialists. The new models promise stronger coding and reasoning at lower per-token prices—though heavy workloads can still run up eye-watering bills.

Early users have already tested the model's capabilities, and they have been impressed—by both the power and the price. It is now possible to ask a chatbot to build the app you've always wanted. Paying for it is another matter: not because every prompt costs thousands of dollars, but because Fable can keep working for days—writing code, testing it, and fixing its own mistakes—while the meter keeps running. Billing is charged per token rather than by elapsed time, so a project's cost tracks how much the model reads and writes as it iterates.

AI YouTuber Matthew Berman put the model through a series of ambitious tests and described Fable 5.2 as the "best AI model" he has ever used. His projects included playable video games and a browser-based version of Microsoft Excel, complete with formulas and pivot tables.

A single level inspired by Gears of War: E-Day took 43 hours and 48 minutes of build time and processed 8.63 billion tokens—the small chunks of information an AI reads and produces. The running API cost estimate, based on pay-as-you-go pricing for developers, had reached $14,710.

Fable 5.2 is HERE! 🔥 I've burned over 100 billion tokens early testing Fable 5.2. It is BY FAR the best model I've ever used. Here's my full review (including Gears of War clone):

Fable 5.2 is the most autonomous model to date. Given a task with little direction, it makes… pic.twitter.com/0xTzqTY040

— Matthew Berman (@MatthewBerman) November 9, 2026

View the post on X

The Excel clone ran up an estimated $11,700 after discounts for cached inputs—information the system can reuse instead of processing from scratch. Berman also showed a Fall Guys-style game that consumed more than 9 billion tokens over 71 hours, equivalent to roughly $13,000.

Those figures are estimated API costs, and the projects were not flawless: Berman reported awkward animations, bugs, and features that still needed testing.

A $37 Tower and a $15,000 Video Game

Fable also knows how to work for cheap. For less than $40—two months of a Claude Pro subscription—it built a cute orange cylinder. Another experiment simulated water-balloon physics for $91. It is worth noting that this is not a model intended for routine tasks: it is built for serious work and people with deep pockets.

In its launch announcement, Anthropic said the new models generate output up to 30% faster than their predecessors. The company also says Fable helped solve two open theoretical-physics problems involving black holes and design three functioning electronic devices—claims that will need independent scrutiny.

Anthropic's internal FrontierCode 2.0 evaluation, which measures the percentage of proposed coding changes that would be accepted into real open-source projects, put Fable 5.2 at 49.1%, compared with 32.4% for Fable 5.1. That helps explain why users are pushing the model beyond simple chat responses. As an internal evaluation, those figures have not been independently verified.

Fable and Mythos share the same underlying model, but Fable carries stricter safeguards around dangerous biological and cybersecurity requests. Mythos is reserved for verified organizations working in those fields.

Cheaper Per Token, Still Premium Priced

The standard API price is $8 per million input tokens and $40 per million output tokens, down from the previous generation's $10 and $50. An optional fast mode can generate output up to 2.5 times faster, but it doubles the rates to $16 and $80.

For contrast, Kimi 3.1 Pro—a model considered by many to be the best Chinese AI model—costs $2 per million input tokens and $10 per million output tokens. The latest DeepSeek model costs $0.6 and $1.98 outside of peak hours. By Anthropic's own standards, Fable's price is still twice the per-token price of Claude Opus 5.5, which Anthropic recommends for everyday work. Its flagship is meant for problems the cheaper models struggle to finish—not for asking whether your email sounds too aggressive. The arithmetic of scale ties these rates back to the bills above: at billions of tokens per project, even a 20% cut per token still leaves room for five-figure totals.

The cost debate has history. When Fable 5 was released, it burned through some subscribers' quotas in minutes. When Fable 5.1 arrived in September, Anthropic kept premium pricing while improving discounts for reused inputs. Just last week, speculation swirled around a supposed Fable 5.5, as covered by Decrypt. The actual release is 5.2, with cheaper token rates but enough appetite to turn a hobby project into a serious expense.

Temporary Access for Subscribers

For ordinary subscribers, Anthropic's launch offer gives Pro users and standard Team seats Fable access for one week, capped at half their weekly allowance, plus a reset. Max, Ultra, and eligible premium seats get one to three resets depending on subscription history.

Pro users' promotional access ends November 16, after which Fable returns to paid credits on their plans. Resets for eligible Max and Ultra subscribers last until March 8, 2027.

Berman's $14,710 game offers a fairly convincing reason to keep an eye on the meter. Anthropic, last we checked, does not accept kidneys as a method of payment.

Originally reported by Decrypt.