Claude Fable 5.2 Is Out. Should Your Team Switch, and Who Pays for It?

Claude Fable 5.2 shipped today, and for most teams the right move is to keep Claude Opus 5.5 as the everyday default and give Fable 5.2 to the people doing long, hard, open-ended technical work. The bigger decisions this week are about seats and admin settings: on Pro and standard Team seats every Fable message is paid from usage credits, and on Enterprise the new model turns itself on for some roles after seven days unless an admin decides first.

Here is what actually shipped, and the short list of things I'd check before Monday's standup turns into "why did our usage spike?"

What Anthropic shipped on November 9

Anthropic released two models at once: Claude Fable 5.2 and Claude Mythos 5.2. They are the same underlying model. Fable 5.2 carries extra cybersecurity and biology safeguards and is open to every paid plan. Mythos 5.2 has no such safeguards and is restricted: security professionals apply through the Cyber Verification Program and life sciences researchers through the Life Sciences Verification Program.

The numbers that matter for planning:

• API price: $8 per million input tokens and $40 per million output tokens, double Opus 5.5's $4 and $20, per the models overview.
• Fast mode: $16 input and $80 output per million, for up to 2.5x faster generation, available in Claude Code and on the Claude Platform among others.
• Speed and efficiency: outputs up to 30% faster than Fable 5.1, and fewer tokens per task at medium effort.
• Context: 1M tokens in, 128K tokens out, same as the rest of the current lineup.
• Effort default: medium in the Claude apps and Claude Code, high on the API.

Anthropic also says that on its FrontierCode 2.0 agentic coding benchmark, Fable 5.2 at high effort roughly matched Opus 5.5 at max effort for about half the cost per task. That is the most useful line in the announcement, and I'll come back to it.

Should your team make Fable 5.2 the default?

No, and Anthropic says the same thing. Its own guidance is to start with Opus 5.5 as the daily driver, move up to Fable or Mythos for "the most ambitious, open-ended technical work", and move down to Sonnet 5.5 or Haiku 5.5 when speed matters more. The docs put it the same way: use Fable 5.2 for demanding reasoning and long-horizon agentic work, "or when your evals on Claude Opus 5.5 at higher effort still fall short."

On team calls, the pattern I see is that people blame the model for problems that are really setup problems: no project context, no CLAUDE.md, a vague first message. A bigger model does not fix that, it just costs more per attempt. So my rule for a team rollout is simple:

1. Default everyone to Opus 5.5. Drafting, research, analysis, Cowork tasks, and most Claude Code sessions.
2. Name the Fable jobs. A multi-hour refactor, a migration, a hard debugging session where Opus has already stalled at high effort. Write the list down so people don't guess.
3. Try the half-cost claim on your own work. If an engineer is running Opus 5.5 at max effort on a long agentic task, run the same task once on Fable 5.2 at high effort and compare quality and usage. Benchmarks are Anthropic's numbers; your repo is yours.

I wrote about the same trade-off when Opus 5.5 became the default in what that change actually costs you, and the cheap end of the lineup in where Haiku 5.5 earns its place.

Which Claude plans include Fable 5.2?

This is where teams get surprised. Per Anthropic's help article on Fable models on your plan, Fable 5.2 is available on every paid plan, but how it is paid for depends on the seat:

• Max, Ultra, premium Team seats and premium seat-based Enterprise seats: Fable is part of the plan. Each person can spend up to 50% of their weekly limits on it. That is not extra allowance: it comes out of the same weekly pool, and Fable uses that pool faster than other models.
• Pro, standard Team seats and standard seat-based Enterprise seats: Fable is not included in usage limits. It runs on pay-as-you-go usage credits from the first message.
• Usage-based Enterprise and the API: billed at standard API rates.

Anthropic also lists limited-time launch offers for eligible paid plans, linked from that help article, so check what your plan gets before you assume the worst.

If your team is mostly on standard seats, the practical question is not "is Fable better" but "who on the team should be spending credits on it." That is a policy decision, and it is easier to make on day one than after the first invoice.

Enterprise admins: the seven-day switch

On Team and Enterprise plans, an owner decides which models people can pick. Enterprise has one detail worth putting on a calendar. Fable 5.2 stays off until an admin enables it, but seven days after release it is added to every role that already has any Fable model enabled, and any default set to a Fable model moves to 5.2, unless an admin opts out first.

From a November 9 release, that lands around November 16. Before then:

• List the roles that have Fable 5 or Fable 5.1 enabled today. Those roles get 5.2 automatically.
• Check defaults. If any role defaults to a Fable model, its people move to 5.2 and its pricing. Decide whether that role should default to Opus 5.5 instead.
• Tell people. A one-paragraph note ("Opus 5.5 is the default; use Fable for these three kinds of work") saves a week of confused Slack threads.
• Update Claude Code. Fable 5.2 needs Claude Code 2.1.321 or later; Cowork needs the latest Claude Desktop.

Why some answers will come from a different model

Fable 5.2's safety classifiers block high-risk cyber work (penetration testing, exploit generation, binary vulnerability analysis) and most research biology tasks. When a cyber request is blocked, Claude Opus 4.8 answers instead; a blocked biology request goes to Claude Opus 5. Anthropic says 5.2 blocks about 55% less often than Fable 5.1.

If your security team or a lab group reports that Fable "got worse" on their work, this is the first thing to check. The verification programs above are the route for legitimate work. Enterprise customers also get Claude Security, which scans codebases with the most capable models, including Mythos 5.2, without joining the Cyber Verification Program.

The same week, Anthropic added a security center, inference hooks and more access controls to Claude Enterprise, several in beta, described in more visibility and control for security teams. If your IT team was waiting on admin controls before widening a rollout, that post is worth forwarding with this one.

A one-week rollout checklist

• Keep Opus 5.5 as the team default.
• Write down the two or three kinds of work that justify Fable 5.2.
• Check seat types: who is on standard seats and will pay in credits?
• Enterprise: review Fable-enabled roles and defaults before about November 16.
• Update Claude Code to 2.1.321 or later on every machine that will use it.
• Run one side-by-side test of Opus 5.5 at max effort against Fable 5.2 at high effort on a real task.
• Brief whoever handles security questions on the model fallback for blocked requests.

If the effort setting itself is new to your team, my earlier post on the effort dial in Claude Opus 5 explains why it often matters more than which model you picked.

Sources: Introducing Claude Fable 5.2 and Claude Mythos 5.2, Models overview, Claude Fable models on your plan, Claude release notes, More visibility and control over Claude Enterprise for security teams.

If you want help setting a model policy and getting your whole team using it well, that is what private team training is for, and if it's just you, 1:1 coaching gets you there faster.