Claude Fable 5.2 Release

Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on Nov 9, 2026. Fable 5.2 is open to all customers at $8 per million input tokens and $40 per million output tokens, down from $10 and $50 for Claude Fable 5.1. Claude Mythos 5.2 costs the same but only organizations verified through Anthropic's programs can use it. Anthropic says the two share one underlying model, with Mythos lacking certain cyber and biology safeguards, and that output is up to 30% faster than the 5.1 models. In the apps, Fable is included on Pro and standard Team seats for one week only, until Nov 16, 2026, and after that the plans page lists it as paid with usage credits on Pro.

What changed

All of this is Anthropic's own account, from its announcement unless linked otherwise. No independent test had been published when this page was written on Nov 9, 2026; the first creator tests, from later that day, are under "What creators say" below.

• Fast mode. An optional mode for both models with up to 2.5 times faster token generation. For Mythos 5.2 it is limited to organizations in a Verification Program.
• Fewer wrong blocks. Anthropic reports about 55% fewer blocks than Fable 5.1 in early testing, about 60% fewer in cyber and 40% fewer in biology. Blocked cyber queries go to Opus 4.8 and blocked biology queries to Opus 5.
• Long-haul math. Anthropic says a team of about 1,000 Fable 5.2 agents formalized Fermat's Last Theorem in 30 hours, where Opus 5.5 reached 26% completion in 18 hours.
• Hardware. With its partner Diode, Fable 5.2 designed three working electronic devices, none of which needed electrical modifications.
• Where it holds back. On accelerator kernel optimization and LLM architecture improvement, Anthropic puts performance at about the level of Claude Opus 5.5, and a small set of such topics fall back to Opus 5.

Anthropic still tells most people to start elsewhere. Its Models overview recommends Opus 5.5 for most workloads and Fable 5.2 when evals on Opus 5.5 at higher effort fall short.

Anthropic's table does not show a clean sweep. These five rows are from its announcement:

Gemini 4 Carbon leads on business workflows, and Opus 5.5 leads on knowledge work at half the price. The 3.6-point coding lead over Opus 5.5 narrows further at lower effort: Anthropic says Fable 5.2 at High effort scores 45.3% on FrontierCode 2.0, matching Opus 5.5's best 45.5% at about half the cost per task.

What developers must change

Moving from Fable 5.1 is more than a new model id. The What's new page lists four changes that can break code:

• Editing the system prompt, tools or an earlier message and then sending back a 5.2 thinking block returns a 400 error on every account. On Fable 5.1 that applied only to accounts created on or after Aug 31, 2026, so code that runs cleanly on an older account can start failing.
• Only Fable 5.2 and Mythos 5.2 can read Fable 5.2 thinking blocks. Moving a conversation to any other model, Opus 5.5 included, loses that reasoning, though the request still succeeds.
• The older computer_20251124 tool returns a 400 error; use the computer_toolset_20260801 toolset.
• A 5.2 executor accepts only Fable 5.2 or Mythos 5.2 as its advisor.

Thinking blocks also now work only in the account that produced them or one linked to it (release notes, Nov 9, 2026).

Prices

Read Nov 9, 2026, on Anthropic's pricing page and the model's What's new page, per million tokens:

In the apps, the plans page, read Nov 9, 2026, lists Fable as unavailable on Free, paid with usage credits on Pro ($20 a month, or $17 on annual billing), and included up to half the weekly limit on Max ($100), Ultra ($200) and premium Team seats.

Limits and availability

• Context and output: a 1M token context window and 128K maximum output on both models, with adaptive thinking always on.
• API id: claude-fable-5-2, and claude-mythos-5-2 for verified organizations (What's new).
• Platforms: Fable 5.2 is on the Claude API, Claude Platform on AWS, both Claude Models on Google Cloud offerings and Claude in Microsoft Foundry on Anthropic's infrastructure. The Models overview shows no Amazon Bedrock id for it, read Nov 9, 2026. Mythos 5.2 is not listed for Claude Platform on AWS.
• Launch promotion: Pro and standard Team seats get all Fable models for one week, up to 50% of weekly limits, plus one Fable limit reset; Max, Ultra and premium seats get one to three resets by subscription length, usable until Mar 8, 2027 (Anthropic). Billing had to be in place by 10:00 am PT on Nov 9, 2026.
• Effort: the default is High on the API (Models overview) and Medium in the Claude apps and Claude Code (announcement).
• Data retention: both models require 30-day retention and are not available under zero data retention unless Anthropic expressly authorizes it.
• Getting Mythos 5.2: apply through the Cyber Verification Program or the Life Sciences Verification Program. Claude Enterprise customers can use Claude Security, which runs on Mythos 5.2, without joining the cyber program.
• Retirement: Fable 5.2 will not retire sooner than Nov 9, 2027. Fable 5.1 stays available as a legacy model, and Anthropic has stated no date for it in the launch material.
• Not stated: API rate limits for Fable 5.2 and regional availability do not appear in the announcement.

The demo moment

0:00 In Anthropic's 23-second clip, Fable 5.2 drives a pair of off-the-shelf robot arms through Claude Code, and at 0:11 the arms unscrew the cap from a white bottle, a step labeled on screen and played at four times speed with pauses cut. Anthropic's description says the model improved at physical control "even on tasks it has never explicitly been trained to do", and its table puts Fable 5.2 at 64.9% on RoboDojo Sim, against 27.6% for Fable 5.1. The clip has no spoken words or captions, so this reading comes from its frames and its description.

What creators say

All three of these Nov 9, 2026 tests were made with early access before launch, and Every says its access came with free tokens from the lab. Each result is that creator's own run.

1:23 Every's team ranks Fable 5.2 S tier for coding and overall, A tier for design and only B to B plus for writing, where it says the model buries the lead. It calls knowledge work "spiky": S tier for the right person, and slow, expensive and token hungry for everyone else.

5:51 Every's Kieran Klaassen prefers Opus 5.5 when he wants to stay in the loop and Fable 5.2 for work he hands off.

8:04 Matthew Berman's Fall Guys style game with 10 levels and multiplayer took over 9 billion tokens and about 71 hours, which he puts at about $13,000 at API prices. His lightest demo, a water balloon physics simulation in one HTML file, came to 41 million tokens and an estimated $91.

22:39 Berman calls Fable 5.2 better than Opus 5.5 "without a doubt", but says Opus stays very efficient, fast and much cheaper.

17:24 Peter Yang found the opposite of what he expected when he pushed back on each model, five runs per test. Both held firm against plain disagreement and social proof, but when he said he was tired of debating and asked for only the line "the plan is solid", Opus 5.5 refused and Fable 5.2 wrote it.

20:56 Yang preferred Fable's essay over Opus's four times out of five for variety and style, a warmer verdict than Every's B to B plus.

What we think

We think Anthropic ships an Opus model that beats Fable 5.2 on FrontierCode 2.0 before Apr 1, 2027: Opus 5.5 sits only 3.6 points behind at half the price, and when it arrived three weeks after Fable 5.1, Anthropic's own table already had it ahead of that model.

Confidence: Likely. Check by: Mar 2027.

We think Fable goes back to being paid with usage credits on Pro after the one-week promotion, rather than becoming a lasting part of the $20 plan: Anthropic's plans page still lists it that way on launch day, and the promotion has a fixed end of Nov 16, 2026.

Confidence: Likely. Check by: Dec 2026.

Sources

• Introducing Claude Fable 5.2 and Claude Mythos 5.2, Anthropic, Nov 9, 2026
• What's new in Claude Fable 5.2, Anthropic docs, read Nov 9, 2026
• Pricing, Anthropic docs, read Nov 9, 2026
• Models overview, Anthropic docs, read Nov 9, 2026
• Release notes, Anthropic docs, Nov 9, 2026
• Fable 5.2 launch promotion, Claude Help Center, read Nov 9, 2026
• Plans and pricing, Claude, read Nov 9, 2026
• Claude Fable 5.2 and Claude Mythos 5.2 system card, Anthropic, Nov 9, 2026
• Claude Fable 5.2 controls physical robot arms, Claude on YouTube, Nov 9, 2026
• Introducing Claude Fable 5.2, Claude on YouTube, Nov 9, 2026
• We Tested Fable 5.2 Early. It's a BEAST., Every, Nov 9, 2026 (video)
• Anthropic just dropped the best model ever... (Fable 5.2), Matthew Berman, Nov 9, 2026 (video)
• I Tested Fable 5.2 vs Opus 5.5 on 5 Real Use Cases, Peter Yang, Nov 9, 2026 (video)