Claude Fable 5.2: $8/$40 per MTok, what breaks from 5.1

On November 9, 2026, Anthropic released Claude Fable 5.2 (claude-fable-5-2), a model it describes as "built for demanding reasoning and long-horizon agentic work." It costs $8 per million input tokens and $40 per million output tokens, down from $10 and $50 on Fable 5.1 (Anthropic's what's new page). It has a 1M-token context window and 128K max output.

Do you need it? Probably not yet. Anthropic's own advice is to start with the latest Opus and "move to Claude Fable 5.2 when your evals on a Claude Opus model at higher effort still don’t meet your requirements." This site is independent and not affiliated with Anthropic, so everything below comes from Anthropic's docs. We did not run Fable 5.2 ourselves.

What Fable 5.2 costs

Source: Anthropic's what's new page, pricing section, checked 2026-11-09.

Fast mode is the optional faster-output setting. Anthropic's announcement says it "speeds up token generation by up to 2.5 times" (announcement). These are API rates, billed per token. They are separate from a Pro, Max or Team subscription, which Claude pricing covers.

Where it sits in the lineup

All four list a 1M context window, 128K max output and a June 2026 knowledge cutoff (models overview for Fable 5.2). Fable 5.2 is twice the input price of Opus 5.5, and four times Sonnet 5.5. For the Opus and Sonnet trade-off, see Claude Opus vs Sonnet.

Claude Mythos 5.2 (claude-mythos-5-2) has the same capabilities, specs and prices. It is available only to "organizations verified through Anthropic’s verification programs."

Use it in Claude Code

Fable 5.2 needs a recent Claude Code. The model configuration docs say it requires v2.1.321 or later. Update, then select it yourself, because no Fable model is the default on any plan:

claude update
claude --model fable

Inside a session, /model fable does the same. The fable alias points at Fable 5.2 outside Claude apps gateway sessions, where you type /model claude-fable-5-2 instead.

Claude Code runs Fable 5.2 at medium effort by default, while the raw API default is high. Effort is the dial for thinking depth and cost: see how to set Claude Code's effort level. Fast mode is also available in Claude Code (Claude Code fast mode).

On plans, Anthropic says Fable "is included for up to 50% of weekly usage limits for Pro plans and standard seats on the Team plan." For how weekly limits work, see Claude usage limits.

What breaks if you call the API from Fable 5.1

If Claude Code, claude.ai or the Agent SDK builds your requests, Anthropic says nothing changes for you. If you write the API calls yourself, four changes from Fable 5.1 can break your code:

1. Editing earlier turns. Changing the system prompt, the tools or an earlier message, then sending back a thinking block, now fails on every account. On Fable 5.1 it only failed on accounts created on or after August 31, 2026.
2. Switching models mid-conversation. Fable 5.2 thinking blocks are read only by Fable 5.2 and Mythos 5.2. Other models drop them, so the reasoning is lost for those turns.
3. Old computer use. The computer_20251124 tool returns an error. Use computer_toolset_20260801.
4. Advisor tool. A Fable 5.2 executor accepts only a Fable 5.2 or Mythos 5.2 advisor.

Two rules apply to every request: adaptive thinking is always on, and forced tool use (tool_choice of any or tool) returns a 400. The code change itself is one line:

model = "claude-fable-5-1"  # Before
model = "claude-fable-5-2"  # After

Anthropic's migration guide has a test for the first change. Run it before you switch.

Where Fable 5.2 can disappoint

Anthropic lists behavior differences that show up with no code change:

• It thinks even when told not to. A prompt line asking for less thinking did not stop it at medium effort and above. At high and above, "the model thinks before almost every reply."
• A turn can end early. On long tasks it sometimes stops with a list of what remains. This happens more at low and medium effort.
• Subagents add cost. In harnesses that invite delegation, Anthropic says subagents "can account for a large part of what a session costs."
• It can refuse. It runs safety classifiers, and a declined request returns HTTP 200 with stop_reason: "refusal".
• Retention. It requires 30-day data retention and is unavailable under zero data retention unless Anthropic authorizes it.

FAQ

How much does Claude Fable 5.2 cost?

$8 per million input tokens and $40 per million output tokens at standard API rates, down from $10 and $50 on Fable 5.1. Batch is $4 and $20. Fast mode is $16 and $80.

Should I switch from Opus 5.5 to Fable 5.2?

Only after testing. Anthropic advises starting with the latest Opus and moving to Fable 5.2 when Opus at higher effort still misses your requirements. Fable costs twice as much per input token, and it thinks before almost every reply at high effort and above.

What is Claude Mythos 5.2?

A model with the same capabilities, specs and prices as Fable 5.2, available only to organizations verified through Anthropic's verification programs, such as the Cyber Verification Program. Everyone else uses Fable 5.2.

How do I use Fable 5.2 in Claude Code?

Run claude update to reach v2.1.321 or later, then start with claude --model fable or type /model fable in a session. No Fable model is the default on any plan, so you must select it explicitly.

Want the short version of the next model launch in your inbox? Subscribe below.

Sources (checked 2026-11-09)

• What's new in Claude Fable 5.2: pricing, breaking changes, behavior differences, availability
• Claude Fable 5.2 overview: release date, specs, lineup comparison
• Claude Platform release notes: the November 9, 2026 entry
• Anthropic's Fable 5.2 and Mythos 5.2 announcement: fast mode, plan usage
• Claude Code model configuration: minimum version, alias, effort default