Models · 7 min read ·
Claude Fable 5.2 in Claude Code: price and setup
Claude Fable 5.2 launched on November 9, 2026. How to select it in Claude Code, what it costs on Pro, Max and the API, the launch offer, and when to use it.
What is Claude Fable 5.2?
Claude Fable 5.2 is Anthropic’s most capable generally available model. It launched on November 9, 2026, together with Claude Mythos 5.2, in Anthropic’s announcement. The two are the same underlying model. Fable carries cybersecurity and biology safeguards, which is why every paid plan can use it; Mythos goes only to organizations Anthropic has verified.
Three things changed from Fable 5.1, according to the announcement. It writes output up to 30% faster. It costs less on the API: $8 per million input tokens and $40 per million output tokens, where Fable 5.1 cost $10 and $50, as VentureBeat reported. And its safeguards interrupt less often: in early testing across claude.ai, Claude Code and Cowork, Anthropic saw about 55% fewer blocked requests than with Fable 5.1.
Every fact below comes from that announcement, Anthropic’s model page, the Claude Code model docs and its help center, read on November 12, 2026. The benchmark figures are Anthropic’s own. We make SwarmPane, which runs Claude Code in panes, so weigh the last section accordingly.
| Fact | Value |
|---|---|
| Released | November 9, 2026 |
| Model ID | claude-fable-5-2 |
| In Claude Code | /model fable, on version 2.1.321 or later |
| Context window | 1 million tokens |
| Longest reply | 128,000 tokens |
| API price | $8 per million input tokens and $40 per million output tokens |
| Fast mode price | $16 and $80 per million tokens |
| Effort levels | low, medium, high, xhigh and max; Claude Code starts at medium |
| Knowledge cutoff | June 2026 |
How do you use Fable 5.2 in Claude Code?
Update first. Fable 5.2 needs Claude Code 2.1.321 or later, so run claude update. Then run /model fable in a session, or start one with claude --model fable. Both are in the model docs.
You have to pick it yourself: the same docs say no Fable model is the default on any plan or provider. /model fable also saves Fable as your default for new sessions. To use it for one session only, open the picker with /model and press s on its row.
If you had Fable 5.1 saved, you may already be on 5.2. On Pro, Max and Ultra plans, a saved claude-fable-5-1 moves to the fable alias the first time you run 2.1.321 or later, and the startup line can show (auto-updated) once. On Team and Enterprise plans the move waits for your organization’s settings. Run /model to see which model you are on.
Then set the effort. Fable 5.2 starts at medium in Claude Code and goes up to max, and /effort changes it. Higher effort reasons more deeply and costs more, so raise it for the hard problem, not for the whole day.
What does Fable 5.2 cost on your plan?
It depends on the plan, and Anthropic’s help article spells it out. On Max and Ultra plans, and on premium seats of Team plans, Fable is part of the plan: you can spend up to 50% of your weekly usage limit on Fable models at no extra cost. Fable uses that limit faster than other Claude models. Past the 50%, you continue on pay-as-you-go usage or switch to another model.
On Pro plans and standard Team seats, Fable is outside the plan’s limits. It runs on pay-as-you-go usage from the first request, and the /model picker marks the Fable row when that is the case. In an interactive session Claude Code asks before it starts billing that way, per the model docs. In a claude -p run it does not ask.
On the API the price is per token: $8 per million input tokens and $40 per million output tokens, with cache reads at $0.20 and half price through the Batch API, per the model page.
| Plan | How Fable 5.2 is billed | Launch offer |
|---|---|---|
| Pro, and standard seats on Team | Pay-as-you-go usage, outside the plan’s limits | Included until November 16, 2026, up to 50% of the weekly limit, plus one Fable reset |
| Max, Ultra, and premium seats on Team | Included, up to 50% of the weekly limit | One to three Fable resets, to use by March 8, 2027 |
| Free | Not available | Not eligible |
| Claude API | Per token: $8 in and $40 out per million | Not a plan offer |
What is the Fable 5.2 launch offer?
Until November 16, 2026 at 11:59 pm Pacific time, Pro plans and standard Team seats get every Fable model inside the plan, up to 50% of the weekly limit, according to Anthropic’s launch offers article. After that 50%, or once the week ends, Fable goes back to pay-as-you-go usage. So this is the cheap week to try it on Pro.
Eligible accounts also get Fable limit resets. Pro and standard Team seats get one, which expires with the week. Max, Ultra and premium seats get one for under six months as a paying subscriber, two for six to twelve months and three for more than twelve, and those expire on March 8, 2027. The account needed paid billing in place before November 9, 2026 at 10:00 am Pacific time.
A reset gives back the usage you spent on Fable models this week and, if your account has a five-hour limit, refills that window. It does not raise your limits. To use one, open Settings > Usage on claude.ai and choose “Reset for free” under Limit resets. You cannot undo it, so keep it for the day you run out.
Is Fable 5.2 better than Opus 5.5 for coding?
On Anthropic’s own benchmarks, yes. In the announcement’s table, Fable 5.2 scores 49.1% on FrontierCode 2.0, which measures whether open-source maintainers would merge a model’s changes, against 45.5% for Opus 5.5 and 42.8% for GPT-6 Astra. On FrontierSWE v2 it scores 69.9%, against 62.3% and 65.5%. These are the vendor’s numbers, each model at its highest effort, so read them as a direction and not a verdict.
The more useful figure is cost per task. Anthropic says Fable 5.2 at high effort scores about the same as Opus 5.5 at its best (45.3% against 45.5%) for roughly half the cost per task. That holds even though Opus 5.5 costs half as much per token, at $4 and $20 per million: Opus 5.5 needs its max effort to get there.
Anthropic still does not recommend Fable for everything. Its advice is to start most coding and knowledge work on Opus 5.5, which it calls its “strongest daily driver”, and to move up to Fable for the most ambitious, open-ended work. The Claude Code docs name what that is: root-cause investigations, outage debugging, architecture decisions, and tasks you would normally break into pieces. They also say to describe the outcome and not the steps, and to skip the reminders to test, because Fable verifies its own work more often.
| Fact | Fable 5.2 | Opus 5.5 |
|---|---|---|
| Select it with | /model fable | /model opus |
| API price per million tokens | $8 in, $40 out | $4 in, $20 out |
| On Pro | Pay-as-you-go usage after the launch week | Inside the plan’s limits |
| FrontierCode 2.0 | 49.1% | 45.5% |
| FrontierSWE v2 | 69.9% | 62.3% |
| Anthropic’s advice | The most ambitious, open-ended work | Start here for most coding |
What is fast mode on Fable 5.2?
Fast mode is the same model answering up to 2.5 times faster at a higher price. Run /fast in Claude Code while you are on Fable 5.2, and a ↯ icon shows beside the prompt while it is on. The fast mode docs say quality and capabilities stay the same.
On Fable 5.2 it costs $16 per million input tokens and $80 per million output tokens, double the standard rate. On every subscription plan, Max included, fast mode bills as pay-as-you-go usage and is not part of the plan’s limits. Use it for live debugging and quick iteration, and turn it off when cost matters more than speed.
What does Fable 5.2 refuse to do?
Fable 5.2 blocks high-risk cybersecurity work, such as penetration testing and exploit generation, and most research biology tasks. You still get an answer, from another model: Opus 4.8 for cybersecurity and Opus 5 for biology, per the announcement. In Claude Code this shows as an automatic model fallback.
Two more limits come from the model page. Fable 5.2 requires 30-day data retention, so it is not available under zero data retention unless Anthropic has authorized it. And thinking is always on: you control its depth with effort, not with a switch.
Use Fable for the hard task and a cheaper model for the rest
Fable uses your weekly limit faster than other Claude models, so the setup that lasts is a split: one session on Fable for the problem that needs it, and the routine work on Opus 5.5, Sonnet 5.5 or Codex beside it. Run multiple Claude Code sessions at once explains how to keep them from colliding, and Claude Max vs Pro how to tell which plan your usage needs.
SwarmPane puts those sessions in one window. Each pane is a real terminal running Claude Code or another CLI, so /model fable and /effort work in a pane as they do in your own terminal. Press s in the /model picker to change one pane only. Agent Grid shows up to 16 panes at once, Attention lists the ones waiting for you first, and Usage shows a row per CLI from its own reports. When Claude Code hits a limit, Limit handoff moves the work to your other account or to Codex, with a brief that you edit first.
SwarmPane runs the agent CLIs and accounts you already have, so Anthropic bills its plan separately. Get SwarmPane. Starter and Pro start with a 5-day trial for $1.