Fable 5 usage: in-plan on Max, credits on Pro — the honest math

Quick summary

• Max / Team Premium: Fable 5 stays in-plan (up to 50% of your weekly limit), permanent per Anthropic.
• Pro / Team Standard: usage credits at $10 / $50 per million tokens — with a one-time $100 credit, then metered. Batch halves the rate; cache hits cut input to $1/M.
• New since July 24: Opus 5 at $5/$25 is included on Pro — for work that doesn't provably need the flagship, it often makes the credits question moot. Run your numbers in the Max-or-credits calculator before buying credits.
• On credits, front-load the genuinely valuable heavy work — big refactors, research syntheses, reviews — and lean on caching and batch to stretch the $100.
• For Max and Team Premium, Anthropic describes the in-plan restoration as permanent; for Pro, the $100 credit is one-time, then $10 / $50 per million tokens.

The exact wording, from Anthropic

If you're here because you pasted the notice into search, this is the line — verbatim from Anthropic's July 18 announcement on its @claudeai account:

“Beginning July 20, Claude Fable 5 will be included in all Max and Team Premium plans, at 50% of limits. Pro and Team Standard users will continue to have access to Fable via usage credits, and will receive a one-time $100 credit.”— @claudeai, July 18, 2026

What changed: the plans split on July 20

Fable 5's all-plans window was scheduled to close after July 7. Anthropic extended it twice — to July 12, then July 19 — and on July 18 announced the lasting arrangement, in effect since July 20. On Max and Team Premium, Fable 5 keeps drawing from your plan's weekly limit (up to 50%) — permanently, per Anthropic. On Pro and Team Standard it bills through usage credits you buy separately, starting with a one-time $100 credit, then $10 / $50 per million tokens. Your plan still covers Opus 4.8, the new Sonnet 5, and the rest of the lineup as before — those don't change.

What the credits actually cost

The usage-credit and API rate is $10 per million input tokens and $50 per million output tokens — twice Opus 4.8's $5 / $25, and the top of Anthropic's price list. A heavy coding run of roughly 50k input plus 15k output tokens works out to about $1.25; do 30 of those a day and you're near $37.50/day, or $1,100+ a month. Two levers cut that: the Batch API halves the rate to $5 / $25 for work that isn't time-sensitive, and cache hits drop input to $1 per million. Honest caveat: Anthropic hasn't published a separate credit-to-dollar conversion beyond the API price list, so treat these published API rates as your firm reference point.

Is this permanent? What Anthropic actually said

Anthropic framed the credits move as a capacity problem, not a pricing strategy — and on July 18 it delivered on the restoration promise, selectively: Max and Team Premium get Fable 5 in-plan permanently from July 20, while Pro and Team Standard stay on usage credits (with a one-time $100 credit). The practical read: if Fable 5 is central to your daily work, the Max upgrade math now deserves a real look; if you're staying on Pro, budget the $100 as a runway, not a subsidy. We track changes on the status page the day they land.

Which work still deserves Fable 5

At $10 / $50, the question isn't 'is Fable 5 the best model' — it usually is — but 'does this task need the best model.' A working rule: pay flagship rates only for judgment-heavy, one-shot-critical work — architecture decisions, gnarly debugging where wrong guesses are expensive, high-stakes review of work that's about to ship, deep research synthesis. Then pull the two levers on every credit you spend: cache your repeated context (a cache hit bills input at $1 per million instead of $10), and push anything that can wait into the Batch API at half rate. Routine implementation, test generation, migrations, and bulk transforms almost never need the most expensive model on the market.

What to use for everything else

The cheaper routes are genuinely good now. Claude Sonnet 5 launched July 1 as the cheaper agentic model — $2 / $10 per million tokens through Aug 31 (then $3 / $15), roughly 40-60% less than the flagship, scoring 63.2% on agentic coding versus Opus 4.8's 69.2%. Open-weight GLM-5.2 (MIT license) runs about one-sixth the cost of Opus and lands close to it on coding. Pick by the task in front of you, not by the logo — many jobs never needed the most expensive model to begin with.

The way forward

Step back and the pattern is clear: in under a month, the frontier has cycled through launch, shutdown, gated return, and premium metering — and GPT-5.6 spent two weeks behind its own government gate before its July 9 public launch. Betting your workflow on any single flagship's availability is the real cost trap. The durable skill is a workflow that splits work across model tiers: the expensive model where judgment matters, cheap models for the volume, and a routing habit that survives the next policy surprise. That split-your-stack discipline is exactly what the Fable 5 Field Guide covers, issue by issue. Separately, the team behind this site is building Coily, a layer that routes each task to a model you can actually afford and runs it for you — in development, not launched.

From the Field Guide

Stop guessing which model to use

Weekly verified workflows for the credits era: which model in which seat, what it costs per task, and copy-ready prompts — plus an 8-runbook Starter Pack you can download once and keep. Every claim checked against published pricing, sources shown.