Claude Fable 5.2 in Claude Code: what it costs and when to switch

Last updated: November 10, 2026

TL;DR

Anthropic released Claude Fable 5.2 on November 9, 2026, and Claude Code needs version 2.1.321 or later to run it. Update, then run /model fable. It costs $8 per million input tokens and $40 per million output tokens, 20% less than Fable 5.1 and still twice Claude Opus 5.5. Cache reads are the exception: $0.20 per million on both models.

• What changed: Fable 5.2 replaces Fable 5.1 as the model the fable alias picks. Anthropic says it writes output up to 30% faster than 5.1.
• What 2.1.321 changes in /model: fable now means Fable 5.2, best now means the same model as opus, and on Pro and Max a saved claude-fable-5-1 moves to the fable alias.
• What it costs to switch mid-session: each model has its own cache, so the first request after /model re-reads the whole conversation at full input price.
• Who should switch: Anthropic still tells most people to start on Opus 5.5 and move up to Fable for open-ended, long-running work.
• This week only: Pro plans and standard Team seats can use Fable for up to 50% of weekly usage limits, per Anthropic's announcement.

What Anthropic released on November 9

Claude Fable 5.2 and Claude Mythos 5.2 are the same underlying model. Anthropic's announcement, dated November 9, 2026, says Fable carries cybersecurity and biology safeguards that let Anthropic offer it to all users. Mythos is for organizations that apply through its Cyber and Life Sciences Verification Programs.

On FrontierCode 2.0, an agentic coding benchmark that checks whether maintainers would merge a model's changes, Fable 5.2 scored 49.1% at Xhigh effort against 45.5% for Opus 5.5 and 32.4% for Fable 5.1. The cost figure matters more for a Claude Code budget: at High effort, Fable 5.2 scored 45.3%, about level with Opus 5.5's best score, "for roughly half the cost per task," according to the same announcement. Those are Anthropic's reported results.

Claude Code runs Fable 5.2 at Medium effort by default, which the announcement says uses fewer tokens than Fable 5.1 for hard tasks.

What Fable 5.2 costs next to Opus 5.5

Fable 5.2 is twice Opus 5.5 on every line of Anthropic's price list except one. Every price below comes from Anthropic's pricing page, read on November 10, 2026, in dollars per million tokens.

Cache reads cost the same on Fable 5.2 and Opus 5.5 because Anthropic prices Fable cache hits at 0.025 times base input and Opus 5.5 cache hits at 0.05 times. Claude Code caches the conversation as it goes, so the cached part of a long session bills at the same rate on either model. New input and every output token still cost twice as much on Fable.

Fast mode on Opus 5.5 costs exactly what standard Fable 5.2 costs, $8 and $40. If a team already pays for Opus fast mode, standard Fable 5.2 runs at the same per-token price.

What changes when you switch models mid-session

Switching models in the middle of a conversation throws away the cache. Anthropic's prompt caching docs say each model has its own cache, so the first request after /model "reads the entire conversation history with no cache hits." Claude Code asks you to confirm the switch while the cache is still warm.

At Fable 5.2's base input price of $8 per million tokens, re-reading a 400,000-token conversation costs $3.20. Written to the cache at the five-minute rate of $10 per million, it's $4.00. The same 400,000 tokens cost $0.08 at the cache-read rate.

Fast mode has the same shape. Claude Code's fast mode docs say the first time you turn it on in a conversation, you pay the full fast-mode uncached input price for the whole context. On Fable 5.2 that's $16 per million tokens, or $6.40 for the same 400,000-token conversation.

Pick the model when you start the session, not halfway through it.

What Claude Code 2.1.321 changes

Version 2.1.321, published to npm on November 9, 2026, is the first that can run Fable 5.2. Anthropic's model configuration docs list three changes to how /model resolves.

Console accounts, third-party providers, and a claude-fable-5-1 set in project, local or managed settings keep Fable 5.1. If you typed best to get the strongest model, you now get Opus 5.5.

A version older than 2.1.321 can't run Fable 5.2. The docs say a request from an older version fails with "Claude Code does not support this model."

The same release carries security fixes worth updating for on their own, per the Claude Code changelog. One stops MCP OAuth requests from sending secrets such as refresh tokens to another host when a server redirected them. Another closes a case where a read-only Bash command ran outside the sandbox without a permission prompt. A third makes deny rules apply to the file that a cp or mv into a directory writes. Version 2.1.322 followed a few hours later and added fast mode on Vertex.

How to switch to Fable 5.2 in Claude Code

1. Update. Run claude update, then claude -v. Anything below 2.1.321 can't run Fable 5.2.
2. Select it at the start of a session. Run /model fable, or set ANTHROPIC_MODEL=fable before you launch. In a Claude apps gateway session, run /model claude-fable-5-2 instead, since fable resolves to Fable 5 there.
3. Pin it if you need the exact version. Aliases move when Anthropic ships a new model. The full ID, claude-fable-5-2, stays put. On Amazon Bedrock, Google Cloud or Microsoft Foundry, set ANTHROPIC_DEFAULT_FABLE_MODEL to your provider's model ID.
4. Check how it bills. Depending on your plan and seat, Fable can bill to usage credits instead of your included limits. The /model picker shows "Requires usage credits" on the Fable row when it does. In non-interactive mode with -p, Claude Code bills those credits without asking.
5. Turn on fast mode only if you need the speed. Run /fast while Fable 5.2 is selected. It needs usage credits turned on for subscription plans, and it stays on across sessions only if you turned it on while using Fable 5.2.

When to stay on Opus 5.5

Anthropic's own advice is to start on Opus 5.5. The announcement says that "for most coding and knowledge tasks," users should start with Opus 5.5, "our strongest daily driver," and move up to Fable or Mythos as the work requires.

The model configuration docs say where Fable earns its price: root-cause investigations, outage debugging and architecture decisions, plus tasks you'd normally split into pieces. They also say Fable checks its own work with less prompting, so reminders to test are usually unnecessary.

Anthropic's High-effort Fable result matched Opus 5.5's best for about half the cost per task. That's the claim to test on your own repository before a team moves its default, since it decides whether Fable at High costs less than Opus at Max for your work. Output tokens cost twice as much on Fable, so a session that writes a lot of code pays that difference on every line.

Pin the model for the whole team

A model a person picks with /model lives in their own ~/.claude/settings.json. A model value in project settings applies to everyone who opens the repository and reapplies at every launch, per the model configuration docs. Administrators can go further with availableModels in managed settings: claude-fable-5 permits Fable 5, 5.1 and 5.2, while claude-fable-5-1 permits 5.1 only.

Settings in the repository are what keep a team's sessions on the same model when an alias moves, which is what 2.1.321 just did to best. Our AGENTS.md setup shows one way to keep an agent's instructions and checks next to the code. For the first Fable release, see our Claude Fable 5 launch post.

Frequently asked questions

What is Claude Fable 5.2?

Claude Fable 5.2 is the Anthropic model released on November 9, 2026, above Opus 5.5 for long, open-ended technical work. It shares its underlying model with Claude Mythos 5.2 but carries cybersecurity and biology safeguards, which is why Anthropic offers it to all users. It costs $8 per million input tokens and $40 per million output tokens.

How much does Claude Fable 5.2 cost?

$8 per million input tokens and $40 per million output tokens, with cache reads at $0.20 per million, per Anthropic's pricing page on November 10, 2026. That's 20% less than Fable 5.1 and twice Opus 5.5, except cache reads, which cost the same as on Opus 5.5. Fast mode costs $16 and $80.

Which Claude Code version supports Fable 5.2?

Version 2.1.321 or later, which reached npm on November 9, 2026. Anthropic's docs say a request for Fable 5.2 from an older version fails with "Claude Code does not support this model." Run claude update, then claude -v to check the version you're on.

How do I switch Claude Code to Fable 5.2?

Run /model fable at the start of a session, or set ANTHROPIC_MODEL=fable before you launch. In a Claude apps gateway session, run /model claude-fable-5-2. Switch at the start: each model has its own cache, so switching mid-session re-reads the whole conversation at full input price.

Is Fable 5.2 better than Opus 5.5 for coding?

On Anthropic's FrontierCode 2.0 benchmark, Fable 5.2 scored 49.1% at Xhigh effort against Opus 5.5's 45.5%. Anthropic still recommends Opus 5.5 as the daily driver for most coding, and Fable for open-ended, long-running work. Those are Anthropic's own reported results.

Why does /model best give me Opus now?

Claude Code 2.1.321 changed the best alias. It now resolves to the same model as opus. Before that version, it used the model fable resolved to, where Fable was available. To get Fable 5.2, run /model fable or use the full ID, claude-fable-5-2.

Does fast mode work with Fable 5.2?

Yes, from Claude Code 2.1.321. It costs $16 per million input tokens and $80 per million output tokens, and on subscription plans it needs usage credits turned on. Turning it on mid-conversation charges the fast-mode uncached input price for the whole context once.

Can Pro users run Fable 5.2 in Claude Code?

Yes. Anthropic's announcement says Fable is included for up to 50% of weekly usage limits on Pro plans and standard Team seats this week. Depending on plan and seat, Fable usage can bill to usage credits, and the /model picker shows "Requires usage credits" when it does.