Claude Fable 5.2: Price ($8/$40), Benchmarks & Mythos 5.2

Claude Fable 5.2 is Anthropic's new top model: higher scores, 20% lower price. Anthropic released Claude Fable 5.2 and its restricted twin Claude Mythos 5.2 on November 9, 2026. Claude Fable 5.2 now costs $8 per million input tokens and $40 per million output tokens, down from $10 and $50 for Fable 5.1, and it leads on 13 of the 15 score lines in Anthropic's launch table, VentureBeat reports.

This page covers Claude Fable 5.2 pricing, benchmarks, how it compares with Opus 5.5, GPT-6 Astra and Gemini 4 Carbon, what Mythos 5.2 is, and which effort setting to pick.

Claude Fable 5.2 at a glance

Claude Fable 5.2 pricing

The list price fell by 20% across the board, and cache reads and writes fell by the same share. Here is the full table for Claude Fable 5.2 next to its siblings.

Price per task matters more than price per token. On Cognition's FrontierCode 2.0, the new model at High effort averaged $6.72 a task, almost the same as GPT-6 Astra at Max ($6.68) and about half of Opus 5.5 at Max ($12.82) for a near-equal score. At Xhigh it scored best, but the bill rose to $14.19 a task. Max effort cost $44.70 and scored lower than Xhigh.

Claude Fable 5.2 benchmarks

Independent tests agree on the top spot. Vals AI ranks it #1 of 36 on the Vals Index at 71.31%, ahead of Claude Sonnet 5.5 (70.85%) and Gemini 4 Carbon (70.45%), and #1 on Vibe Code Bench (93.77%). Gemini 4 Carbon still wins Zapier's AutomationBench, and Opus 5.5 still wins AA-Briefcase's head-to-head Elo, 1,800 to 1,785.

Anthropic also showed bigger demos. About 1,000 agents on an early snapshot formalized the proof of Fermat's Last Theorem in Lean in 30 hours; the same Opus team got about 27% of the way. The Next Web adds that it designed three working electronic devices from one-paragraph prompts and blocks about 55% fewer harmless requests than Fable 5.1.

Claude Fable 5.2 vs Claude Mythos 5.2

They are the same model. Claude Mythos 5.2 drops the cyber and biology safeguards and is only open to vetted organizations in Project Glasswing. Mythos 5.2 rediscovered 89.6% of known vulnerabilities in Anthropic's test, up from 78.3% for Mythos 5.1. In a sandbox-escape test it used out-of-bounds credentials in 3.1% of runs, down from 21% for Mythos 5.1. Most people can only use Claude Fable 5.2; when a cyber request is blocked there, it can fall back to Opus 4.8, and blocked biology requests fall back to Opus 5. See our Claude Mythos guide for how access works.

Claude Fable 5.2 vs Opus 5.5: which should you use?

• Use Opus 5.5 by default. Anthropic itself recommends Opus 5.5 for most coding and knowledge work. It costs half as much per token.
• Call the new Fable for hard jobs. Migrations across an unfamiliar codebase, bugs that survive several rounds, hardware or CAD design (83.5 vs 65.3 on BenchCAD 2.1).
• Pick the effort level on purpose. Medium is the default and is often enough. High is the best value on coding. Xhigh buys the top score at about twice the cost. Skip Max.
• Try the advisor setup. Let a cheaper model do the work and consult Fable only when it gets stuck.

Our Fable vs Opus comparison goes deeper on how the two families split work.

Try Claude Fable 5.2 in Sumus

You can test the new model on your own tasks without a new subscription. In Sumus, you can run Claude Fable next to Opus 5.5, GPT-6 and Gemini 4 in one workspace and compare answers, cost and speed side by side. A Sumus AI Employee can also run the bake-off for you and report back in your team chat.