Claude Fable 5.2 is out: cheaper than 5.1, in Copilot, and still not your daily model

What happened

Anthropic released Claude Fable 5.2 and Mythos 5.2 on November 9, 2026. They're one model with two sets of safeguards. Fable 5.2 is the general release, on paid plans and the API, and it blocks high-risk cyber and biology work; Mythos 5.2 drops most of those blocks and goes only to organizations accepted into Anthropic's cyber and life-sciences verification programs.

The same day, GitHub made Fable 5.2 generally available in Copilot.

Price and speed

The Batch API halves both rates, per Anthropic's pricing docs. The API model ID is claude-fable-5-2, and Anthropic says it writes output up to 30% faster than 5.1.

Who gets it on which plan

• Free: not included.
• Pro ($20): Fable is in the model picker, but it draws on usage credits you buy on top of the subscription, per the plans page.
• Max ($100) and Ultra ($200): included, up to half of your weekly limit.
• Team: premium seats get it at 50% of weekly limits; standard seats aren't listed.
• Enterprise: included.
• GitHub Copilot: Pro+, Max, Business and Enterprise, billed at Anthropic's list price under usage-based billing. On Business and Enterprise an admin has to switch the Fable 5.2 policy on first; it's off by default.

Paid subscribers also get one to three Fable usage resets as a launch promotion, depending on how long they've subscribed, per the release post.

The numbers Anthropic gives

Outside coding, Fable 5.2 scored 74.2% on Humanity's Last Exam with tools, against 67.7% for Opus 5.5, and 89.3% on OSWorld 2.1, a test of operating a computer. All of these are Anthropic's own runs.

What Anthropic held back

Requests the cyber or biology classifiers block don't fail: they're answered by an older model, Opus 4.8 for cyber and Opus 5 for biology, and Anthropic says 5.2 blocks about 55% less often than 5.1. Anthropic also held back training in two areas, optimizing accelerator kernels and improving LLM architectures, so Fable 5.2 is no better than Opus 5.5 at those. Agentic coding and general ML research weren't capped.

What it means for you

1. Daily work: stay on Opus 5.5. Anthropic's release post calls Opus 5.5 its strongest daily driver, and Opus costs half as much per token. Prompts from the Opus 5.5 prompting guide carry over.
2. Long agentic coding runs: try Fable 5.2 on a task where Opus 5.5 stalled. The 7.6-point gap on FrontierSWE is the case for paying double.
3. Already on Fable 5.1 through the API: switch the model ID to claude-fable-5-2.

Plans and limits are in our Claude pricing guide; to use Fable inside your editor, see GitHub Copilot's plans.

Try it

You're picking up a coding task another model didn't finish.
Goal: [what the change should do, in one or two sentences]
What was tried and where it stopped: [paste the last attempt or the error]
Constraints: [language, framework, files you may not touch, tests that must pass]

First, list what you think went wrong in the last attempt, in at most five lines.
Then make a plan of no more than seven steps and wait for my OK before you change any file.
After each step, run the tests and tell me what changed and what's left.

I pay $4 / $20 per 1M tokens for Opus 5.5 and $8 / $40 for Fable 5.2.
Here's a task I run often: [describe it, with a typical input size in words or tokens].
1) Estimate tokens in and out per run for each model, and the cost of 100 runs.
2) Where, if anywhere, would a stronger model change the result?
3) Recommend one model and say what result would make you change your mind.
If you can't tell without seeing an example, ask me for one.

Sources

• Anthropic, Claude Fable 5.2 and Mythos 5.2, November 9, 2026
• Anthropic, API pricing, checked November 10, 2026
• Anthropic, Claude plans, checked November 10, 2026
• GitHub, Claude Fable 5.2 in GitHub Copilot, November 9, 2026