~/guides/FABLE.md
Claude Fable 5.2 vs Opus 5.5: official scores, price, and the two commands to switch in Claude Code
Anthropic's table: Fable 5.2 beats Opus 5.5 on coding at twice the token price, about half the cost per task on High. Switch with /model fable; plus Mythos 5.2.
Watch the reel on InstagramComment #FABLE to get this guide
Anthropic released Claude Fable 5.2 and Mythos 5.2 on 9 November 2026. This is the full version of the FABLE reel: how far it is ahead of Opus 5.5 on Anthropic's own table, how to read the price so it does not scare you, how to switch Claude Code to it (two commands you can copy), and why Mythos 5.2 needs an application. Every number was checked against Anthropic's own pages on 12 November.
01The result first: Anthropic's own table, Fable 5.2 beats Opus 5.5 on coding

On both coding rows Fable 5.2 scores above Opus 5.5: FrontierCode 2.0 is 49.1% vs 45.5%, FrontierSWE v2 is 69.9% vs 62.3%. FrontierCode measures whether the maintainers of open-source repositories would merge the model's code changes, which is close to what we ask Claude Code to do every day.
- The official scores: which rows it wins and which it does not.
- The price: twice Opus per token, yet in Anthropic's own test it can cost less per task.
- The two commands to switch Claude Code to Fable 5.2, and the plan limits to check first.
- When to use Fable, when Opus is enough, and what Mythos 5.2 is.
02Price: twice the unit price, not always twice the bill

API prices per million tokens: Fable 5.2 is $8 input and $40 output; Opus 5.5 is $4 and $20, exactly half. The launch page's "lower price" and "up to 30% faster" compare Fable 5.2 with Fable 5.1 ($10 / $50), not with Opus.

The unit price is not the whole bill. In Anthropic's early testing on FrontierCode 2.0, Fable 5.2 on High effort (45.3%) scored about level with Opus 5.5's best score on Max (45.5%), for roughly half the cost per task. On Xhigh it reaches 49.1%, the highest of any released model Anthropic tested.
03Switch Claude Code to Fable 5.2: two lines
No Fable model is the default on any plan, so you pick it yourself. The docs say Fable 5.2 needs Claude Code v2.1.321 or later, so update first.
claude update
/model fable
claude --model fable

- Type /model with no argument to open the picker, which lists each model and (on an Anthropic account) its price.
- A model picked with /model is saved as your default; to use it for this session only, press s on its row in the picker.
- In Claude apps gateway sessions, the docs say to type /model claude-fable-5-2 instead.
- On the API, the model name is claude-fable-5-2.
04When to use Fable, and when Opus is enough
Anthropic's own advice: for most coding and knowledge tasks, start with Opus 5.5 (it calls it "our strongest daily driver"), move up to Fable or Mythos when the work is big, open-ended or long, and down to Sonnet 5.5 or Haiku 5.5 when it is simple.
The Claude Code docs also say how to work with Fable: describe the outcome, not the steps; hand it ambiguous problems such as root-cause investigations, outage debugging and architecture decisions; skip the reminders to test, because it verifies its own work; and give it work you would normally split into pieces. Here is that as one prompt you can paste:
The result I want: [e.g. checkout no longer double-charges on mobile, with tests that prove it]. Context: [what the project is, the framework, the relevant folders]. Investigate the cause yourself before changing anything, verify your work when you are done, then tell me which files you changed and what you are still unsure about.
05What Mythos 5.2 is

Fable 5.2 and Mythos 5.2 are the same underlying model. The difference is the safeguards: Fable has strong cybersecurity and biology safeguards, so it is available to everyone; Mythos comes without some of them, is Anthropic's most capable model for cybersecurity work, and is only for vetted organizations. Its API price is the same as Fable 5.2 ($8 / $40 per million tokens).
- Security teams: apply through the Cyber Verification Program.
- Life sciences research: apply through the Life Sciences Verification Program.
- Claude Enterprise customers: no program needed; Claude Security (codebase vulnerability scanning) already uses Mythos 5.2.
06Good to know
- Every score and price in the reel and here comes from Anthropic's own pages, checked on 12 Nov 2026; the scores are Anthropic's own tests and your project may differ.
- "About half the cost per task" is Anthropic's early test on FrontierCode 2.0 (Fable on High vs Opus on Max), not a promise that every task is cheaper.
- "Up to 30% faster" and "lower price" compare with Fable 5.1; per token, Fable 5.2 still costs twice Opus 5.5.
- Before switching, check whether the /model picker says "Requires usage credits", or you may start spending credits.
- Security testing and biology-adjacent projects often get rerouted away from Fable. That is expected routing, not a problem with your account.
Sources
What should the next reel cover?
I test Claude and AI tools every day and turn the useful ones into reels. Follow @vivianlin.tech to catch each new guide.