Claude Fable 5.2: better at code, cheaper, but not for everything
Read in : Français

Contents
Claude Fable 5.2, released by Anthropic on November 9, 2026, is stronger at code and cheaper than the model it replaces, and it was available in Claude Code from day one. But Anthropic's own advice is clear: start with Claude Opus 5.5 for most coding, and step up to Fable for the most ambitious tasks. Here's what changed, how much the benchmark claims are worth and how to pick a model without blowing your budget.
What did Anthropic release on November 9, 2026?
Anthropic shipped two versions of one model: Fable 5.2, open to all customers, and Mythos 5.2, which drops some cybersecurity and biology safeguards and is limited to vetted organizations. For a vibe coder, only Fable 5.2 matters.
Key points from the official announcement and docs:
- API pricing: $8 per million input tokens and $40 per million output tokens, down from $10 and $50 for Fable 5.1.
- Fast mode: up to 2.5 times faster, at $16 and $80 per million tokens. It's available in Claude Code.
- Speed: outputs up to 30% faster than Fable 5.1.
- Default effort: "medium" in the Claude apps and Claude Code, using fewer tokens than Fable 5.1 at the same level.
- Context: 1 million tokens, with up to 128,000 output tokens.
- API ID:
claude-fable-5-2. - Fewer refusals: about 55% fewer blocks on legitimate requests than Fable 5.1.
At launch, Anthropic said Fable can use up to 50% of weekly usage limits on Pro plans and standard Team seats, and that eligible subscribers get one to three Fable usage resets depending on how long they've subscribed.
How much are the benchmark numbers worth?
On Anthropic's published tests, Fable 5.2 beats Opus 5.5 and rival models at agentic coding. These are the vendor's own scores, not an independent lab's.
| Benchmark (agentic coding) | Fable 5.2 | Fable 5.1 | Opus 5.5 | GPT-6 Astra | Gemini 4 Carbon |
|---|---|---|---|---|---|
| FrontierCode 2.0 | 49.1% (xhigh effort) | 32.4% (xhigh) | 45.5% | 42.8% | 41.1% (medium) |
| FrontierSWE v2 | 69.9% | 56.3% | 62.3% | 65.5% | 61.2% |
Two caveats most coverage skips:
- At "high" effort, Fable 5.2 scores about 45.3% on FrontierCode, almost exactly Opus 5.5's 45.5%, at roughly half the cost per task according to Anthropic. The big gap only shows up at the top effort level.
- On AutomationBench (business workflows), Gemini 4 Carbon stays ahead at 53.9% versus 51.4% for Fable 5.2.
Anthropic also says more than 90% of changes merged into its own codebase are now written by Claude, with human guidance. That's an internal figure no one outside can verify.
Why does this matter if you vibe code?
The main message isn't "switch to Fable." It's "match the model to the task." Anthropic's docs recommend Opus 5.5 for most workloads and Fable 5.2 for demanding reasoning and long-horizon agentic work.
It costs twice as much as Opus 5.5
Fable 5.2 costs double Opus 5.5 on the API ($4 and $20 per million tokens). If you work with an API key, here's an illustrative calculation for a session that reads 2 million tokens and writes 200,000, without caching or fast mode:
| Model | Input (2M) | Output (200k) | Total |
|---|---|---|---|
| Fable 5.2 | $16 | $8 | $24 |
| Opus 5.5 | $8 | $4 | $12 |
| Sonnet 5.5 | $4 | $2 | $6 |
Calculated from the API prices in Anthropic's documentation, not from a real measurement. Fast mode doubles Fable's price.
On a subscription, the cost shows up as usage limits: every Fable request eats your allowance faster. For the bigger picture, read what a vibe-coded app really costs.
A quick decision tree
- Fixing a bug, adding a screen, changing a form: Sonnet 5.5 or Opus 5.5 is enough.
- Starting a feature that touches several files and the database: Opus 5.5, with a plan you approve before any code.
- Migrating an architecture, rescuing an app that got out of hand, or the agent has been going in circles for an hour: that's where Fable 5.2 earns its price.
- Touching security or payments: no model replaces review. Go through the vibe coding security checklist.
To compare with OpenAI and Google models, see our guide on which AI model is best for coding and the Cursor vs Claude Code vs Codex comparison.
What to do next
In Claude Code, switching models takes one command. The official docs list these options:
/model fable → switch to Fable and save it as default
/model opus → go back to Opus
/effort high → set the effort level
/effort status → show the current setting
In the picker that opens when you type /model with no argument, pressing s applies the choice to the current session only instead of saving it as your default. That's the right move for a one-off test. You can also run claude --model fable for a single session.
Here's the approach we recommend to stay in control:
- Keep Opus 5.5 as your default and switch to Fable only for a specific task.
- Get a plan first with the request below, and approve it before any code:
Before changing anything, analyze the project and propose a plan:
- which files you'll change and why
- the risks (data, security, regressions)
- the tests to run to verify the result
Wait for my approval before writing code.
- Commit before you unleash Fable on a big job, so you can roll back: see Git for vibe coders.
- Test the result as you would with any model: testing AI-generated code.
- Switch back to Opus once the heavy lifting is done.
If you're new to the tool, start with our Claude Code beginner's guide.
Frequently asked questions
Frequently asked questions
What's the difference between Claude Fable 5.2 and Opus 5.5?
Fable 5.2 is Anthropic's most capable generally available model, built for demanding reasoning and long agentic tasks. Anthropic recommends Opus 5.5 for most coding and knowledge work. Fable 5.2 costs twice as much on the API: $8 and $40 per million tokens, versus $4 and $20 for Opus 5.5.
Is Claude Fable 5.2 available in Claude Code?
Yes. Fable 5.2 has been available in Claude Code since its November 9, 2026 launch. Select it with /model fable or start a session with claude --model fable. The default effort level in Claude Code is "medium," and the faster, more expensive fast mode is available there too.
How much does Claude Fable 5.2 cost?
On the API, Fable 5.2 costs $8 per million input tokens and $40 per million output tokens, less than Fable 5.1's $10 and $50. Fast mode costs $16 and $80. On subscriptions, Anthropic said at launch that Fable can use up to 50% of weekly limits on Pro plans and standard Team seats.
Should I switch all my projects to Fable 5.2?
No. Anthropic itself recommends starting with Opus 5.5 for most coding tasks. Fable 5.2 pulls ahead mainly at the highest effort level and on long jobs such as migrations, rewrites and stubborn bugs. For small edits it burns through your allowance or budget faster with no visible benefit.
What is Claude Mythos 5.2?
Mythos 5.2 is the same model as Fable 5.2, minus some cybersecurity and biology safeguards. It's limited to organizations approved through Anthropic's verification programs and powers Claude Security for Enterprise customers. An individual vibe coder doesn't have access to it and doesn't need it.


