Claude Fable 5.1 vs Sonnet 5: Which Model for Which Job

Run on the included default and escalate to Fable 5.1 only when the problem is genuinely hard. That’s the routing rule, because Fable 5.1 is metered by usage credits even on paid plans. What changed on September 22, 2026: on Pro, Max, and Team the included default is now Opus 5.5, not Sonnet 5. Sonnet 5 is still the Free default and the cheapest API tier, so for API work it remains where you start. Model choice is a budgeting decision, not just a quality one.

The lineup, plainly

Model	What it is	Cost signal
Sonnet 5	The workhorse. Default on Free; the cheap API tier.	Included; API $2/M in, $10/M out — permanent since Aug 11, 2026
Opus 5.5	The value rung. Fable-5.1-level on most work, per Anthropic; replaced Opus 5 on Sep 22, 2026. Default on Pro, Max, and Team.	API $4/M in, $20/M out — 40% of Fable 5.1’s rate
Fable 5.1	The frontier. Released Sep 1, 2026, superseding Fable 5 — same base price, cheaper cache reads.	Usage credits on plans; API $10/M in, $50/M out, cache reads $0.25/M
Haiku 5.5	Fast and cheap. Released Oct 7, 2026, replacing Haiku 4.5.	API $0.10/M in, $0.50/M out up to 100K tokens; high-volume work

Jobs for Sonnet 5 (the Free default and the API workhorse)

Drafting, editing, summarizing, everyday coding, brainstorming, email, scripts, research synthesis. For API work, if you’re unsure which model a task needs, it needs Sonnet 5 at $2/$10. On a paid plan you’ll land on Opus 5.5 by default; pick Sonnet 5 in the model menu when you want faster turns on light work — the quality bar of “default” moved up a full generation this year, and most operators haven’t recalibrated.

Jobs for Fable 5.1 (the escalation)

• Multi-constraint reasoning — legal review, architecture decisions, anything where missing one interaction between requirements is the failure mode
• Hard debugging — the bug Sonnet went in circles on
• High-stakes single documents — the proposal or analysis where a 10% quality gain has dollar value
• Long agentic runs — complex multi-step work where a smarter plan saves more credits than it costs

The tell that you should escalate: you’ve re-prompted Sonnet 5 three times on the same problem. Three retries on the workhorse costs more attention than one clean pass on the frontier model.

Where Opus 5.5 fits (as of September 22, 2026)

Opus 5.5 is the rung between the two. Anthropic’s own framing is Fable-5.1-level performance on most work at 40% of the API price, with output over 30% faster than Opus 5. Two jobs it takes cleanly:

• API work at volume where Fable 5.1’s $10/$50 doesn’t survive the spreadsheet but Sonnet 5 keeps missing. $4/$20 is the middle price for the middle problem.
• Long agentic runs on a budget where the task is hard but not ceiling-hard — the speed and cheaper cache reads ($0.20/M) compound over many steps.

On plans, the practical rule is simpler: Opus 5.5 is the default on Pro, Max, and Team, and it’s the strongest model you can reach without spending usage credits on Fable. Fable 5.1 stays the ceiling; Opus 5.5 is what you reach for before paying ceiling prices — and since it now matches Fable on most work, the ceiling is needed less often.

The credit discipline

Fable 5.1’s metering means an undisciplined “always use the best” habit gets expensive. Run it like a specialist on retainer, not a daily hire:

1. Start every task on the included default: Opus 5.5 on a paid plan, Sonnet 5 on Free or when you’re paying API rates.
2. Escalate on evidence (the three-retry tell), not on vibes.
3. Escalate one rung at a time — Sonnet 5, then Opus 5.5, then Fable 5.1. Most tasks that beat Sonnet stop at Opus.
4. For API work, cache aggressively — Fable 5.1 cut cache-read pricing 75% to $0.25/M, on top of the standing cache discount, so a well-cached agentic session costs meaningfully less than the sticker rate suggests.

Tools need jobs. Models do too. Sonnet 5 has the job called “everything”; Opus 5.5 has the job called “everything, but bigger and harder”; Fable 5.1 has the job called “the thing everything couldn’t do.”