Fable 5.2 in Claude Code: up to 30% faster for 20% less

TL;DR Run claude update to v2.1.321, then /model fable. Fable 5.2 drops from $10/$50 to $8/$40 per million tokens. On Pro it is included until November 16 (half your weekly limit); after that it draws on usage credits.

Fable 5.2 has been the model behind the fable alias in Claude Code since November 9, 2026, but only from v2.1.321 on. Below that version, /model fable still gives you Fable 5.1 at the old price, with no warning. It is still the tier above Opus, with two changes you notice on day one: output is up to 30% faster, and every line of the price list drops 20%.

What changes from Fable 5.1

Cache reads stay at 2.5% of input, the cut that arrived with Fable 5.1. Because all four lines drop together, a long session costs the same 20% less as a short one.

On capability, Anthropic reports 49.1% on FrontierCode 2.0, against 32.4% for Fable 5.1 and 45.5% for Opus 5.5. The number that matters day to day is a different one: at high, Fable 5.2 matches Opus 5.5 at max (45.3% against 45.5%) for roughly half the cost per task.

Check which Fable is answering

The alias is spelled the same on both versions, so the number from claude --version is what decides. To see it in a real response, ask for JSON and read which model billed it:

claude -p "Reply with the single word ok" --model fable --output-format json | jq -c '.modelUsage | keys'

Output on v2.1.319:

["claude-fable-5-1"]

Output on v2.1.321:

["claude-fable-5-2"]

That same minimal request cost $0.184 on Fable 5.1 and $0.141 on Fable 5.2 at list price. If Fable draws on usage credits on your plan, with -p Claude Code bills it without asking, so for a check that costs nothing, the version number is enough.

How to move to Fable 5.2

1. Update Claude Code

claude update
claude --version

You need 2.1.321 or later. If your release channel is stable, or you installed the claude-code Homebrew cask, today you land below it: that channel serves versions about a week old. Switch to latest from /config, or with "autoUpdatesChannel": "latest" in ~/.claude/settings.json.

2. Select the model

/model fable

The choice is saved to your user settings, and later sessions start there. On Pro and Max, a saved claude-fable-5-1 moves to the fable alias by itself on the first run of v2.1.321. To stay on the previous model, select it by ID: /model claude-fable-5-1.

3. Leave effort at medium and raise it only for long work

Fable 5.2 starts at medium, unless you had already set a level for it. Anthropic recommends that level for interactive coding and keeps high for hard work and long unattended runs:

/effort high

If you switch models often, set each model's effort once.

Can't update yet? /model claude-fable-5-2 also answers on v2.1.319, but Claude Code doesn't recognize the model: it logs [claude-code:unrecognized_model] and can't work out its cost. Fine as a stopgap, not as a setup.

What your plan includes

On Pro, once that week ends or you pass the 50%, Fable goes back to usage credits, if you have them turned on. A reset restores what you used; it doesn't raise the limit. Here is where you see what you have left. Which sessions bill without asking, and how to cap the spend: Fable 5.2 on usage credits on Pro.

What you will notice in use

• A turn can end before the work is done. On long tasks it sometimes closes with a list of what remains or an offer to continue, more often at low and medium. For long unattended work, go up to high (what else to do when it stops).
• Telling it to think less no longer works. A line such as "answer directly without deliberating" doesn't stop the thinking at medium and above. If easy requests take too long, lower the effort instead of adding instructions.
• Fewer reroutes to Opus. Anthropic measures about 55% fewer blocks than on Fable 5.1 (60% fewer in cybersecurity, 40% fewer in biology). When one happens, the session continues on Opus until you switch back with /model.
• /fast has its own price. Up to 2.5 times faster at $16/$80, always from usage credits even with plan usage left. Turn it on at the start of a session: the first time, you pay for the whole context uncached.

To decide when it pays off against Opus and Sonnet, this is how I split the models.

Official docs: Model configuration · Fable 5.2 announcement · Launch promotion

Requirements: Claude Code v2.1.321 or later and a paid plan (Pro, Max, Ultra, Team or Enterprise) or API access.