Anthropic releases Claude Fable 5.2 to all paid users

Overview

Updated 54 minutes ago

Anthropic opened Claude Fable 5.2 to every paid subscriber on November 9, 2026. The model shares its underlying system with Claude Mythos 5.2, a version Anthropic limits to vetted cybersecurity and biology researchers.

Anthropic says Fable 5.2 beats its own Opus 5.5 on agentic coding, physics and robot control. It costs $8 per million input tokens and $40 per million output tokens, and Fortune reported the price undercuts OpenAI's GPT-6 Astra.

Why it matters

Frontier labs now ship capability jumps every few months, and each one resets what a single engineer or researcher can attempt.

Questions about this story

Free account needed to ask — your question is kept and asked for you right after sign-up. Answers are public.

No questions yet — be the first to ask.

Key Indicators

Input price per million tokens

Standard rate for Fable 5.2 on the Claude Platform.

$40

Output price per million tokens

Standard rate; fast mode doubles both rates to $16 and $80.

49.1%

FrontierCode 2.0 score

Agentic coding benchmark, run at extra-high effort.

69.9%

FrontierSWE v2 score

Software engineering benchmark; Opus 5.5 scored 62.3%.

64.9%

RoboDojo simulated robot score

Two-arm robot control from camera images, at maximum effort.

Context window in tokens

Maximum output is 128,000 tokens.

30 hours

Time to formalize Fermat's Last Theorem

A team of 1,000 Fable 5.2 agents completed the proof formalization.

Voices

Curated perspectives — historical figures and your fellow readers.

People Involved

Dario Amodei

Co-founder and chief executive, Anthropic

Leading Anthropic through the Fable 5.2 launch

Organizations Involved

Anthropic

AI company

Released Claude Fable 5.2 and Claude Mythos 5.2

Anthropic builds the Claude family of AI models and released Fable 5.2 and Mythos 5.2 on November 9, 2026.

OpenAI

Artificial intelligence company

Maker of GPT-6 Astra, the comparison model in Anthropic's launch

OpenAI released GPT-6 Astra in early September 2026, and Anthropic benchmarked Fable 5.2 against it.

Google DeepMind

AI company

Maker of Gemini 4 Carbon, another comparison model

Google DeepMind's Gemini 4 Carbon appears in Anthropic's benchmark table for Fable 5.2.

Diode

Electronics design and manufacturing company

Built the circuit boards for three devices Fable 5.2 designed

Diode manufactures circuit boards and worked with Anthropic on a test of Fable 5.2's hardware engineering.

Timeline

September 2026 November 2026

4 events Latest: Yesterday

Tap a bar to jump to that date

November 2026
Anthropic releases Claude Fable 5.2 and Mythos 5.2
Latest Product launch

Fable 5.2 goes to all paid users; Mythos 5.2 stays behind verification programs.
Yesterday
Fable 5.2 formalizes Fermat's Last Theorem in 30 hours
Research result

A team of 1,000 agents completed the proof formalization with minimal human input.
Yesterday
Anthropic adds launch-week Fable usage for paid plans
Promotion

Eligible subscribers get one to three Fable resets and up to 50% of weekly limits.
Yesterday
September 2026
OpenAI ships GPT-6 Astra
Product launch

OpenAI released the model Anthropic later benchmarked Fable 5.2 against.
September 2026

Scenarios

OpenAI cuts GPT-6 Astra prices to answer Fable 5.2

Possible Resolves by Feb 28, 2027

Discussed by: Fortune and CNBC, which framed the launch as a price challenge to OpenAI

Anthropic set Fable 5.2 at $8 and $40 per million tokens. If OpenAI matches or goes lower, the price gap Fortune described closes and buyers gain leverage. A cut would show up on OpenAI's public pricing page.

Anthropic files for an initial public offering

Possible Resolves by May 31, 2027

Discussed by: CNBC and Hindustan Times, which reported the launch came ahead of an IPO

Both outlets tied the release to a coming public listing. A filing with the US Securities and Exchange Commission would make it official. Until then, the IPO talk is reporting, not a filed plan.

Mythos stays behind verification programs through 2027

Likely Resolves by Q2 2027

Discussed by: Anthropic's launch post and system card, which describe gated access

Anthropic says Mythos is available only through its Cyber Verification and Life Sciences programs. General release would require a policy change the company has not announced.

Historical Context

3 moments from history that rhyme with this story — and how they unfolded.

March 2023

GPT-4 release (2023)

OpenAI released GPT-4, a model that jumped past GPT-3.5 on standardized exams, coding tests and image understanding. The launch came with a benchmark table comparing it to named rivals.

Then

ChatGPT Plus subscribers got access first, with an API waitlist behind them.

Now

The benchmark-table launch became the standard format for frontier model releases.

Why this matters now

Fable 5.2's launch follows the same script: a benchmark table, named competitors and a same-day rollout to paying users.

January 2025

DeepSeek-R1 price shock (2025)

Chinese lab DeepSeek released a reasoning model and published API prices far below US frontier models. US AI stocks fell as investors recalculated what inference should cost.

Then

Incumbents faced immediate questions about their own per-token pricing.

Now

Price became a headline number in every frontier model launch, not a footnote.

Why this matters now

Anthropic led its Fable 5.2 announcement with price, and Fortune framed it against GPT-6 Astra. Pricing is now part of the product.

November 2020

AlphaFold 2 at CASP14 (2020)

DeepMind's AlphaFold 2 predicted protein structures in a biennial assessment with accuracy that rivaled experimental methods. Specialists in the field judged the results.

Then

Researchers called it a breakthrough in protein structure prediction.

Now

It led to public structure databases and made AI-assisted science a mainstream claim.

Why this matters now

Anthropic now makes a similar claim: Fable 5.2 solved two open problems in black hole physics and formalized Fermat's Last Theorem. The test is whether specialists accept the results.