Claude Fable 5 Free Access Ends July 19: What to Do

The short answer

Claude Fable 5's included access ends on July 19, 2026 at 11:59:59 PM PT. If you do nothing, Fable 5 drops out of your plan's included usage and you keep working on Sonnet 5 and the other included models at no extra cost. If you want to keep using Fable 5 after Sunday, you enable prepaid usage credits, billed at $10 per million input tokens and $50 per million output tokens.

That date is confirmed on Anthropic's own support page, not just the trade press. This is the third deadline in this saga (July 7, then July 12, then July 19), so if you read an article quoting an earlier date, it is stale. Everything below is verified against Anthropic's published documentation as of July 17.

What exactly ends on July 19?

Two things end at the same moment, per the support article above:

1. Included Fable 5 access. Right now, every Pro, Max, Team, and premium seat-based Enterprise plan can use Fable 5 for up to 50% of its weekly usage limits at no extra cost. Fable 5 draws from the same weekly pool as your other models, so heavy Sonnet use eats into the Fable allowance and vice versa. After July 19, Fable 5 is no longer part of that pool at all.
2. The 50% Claude Code limit boost. Anthropic raised Claude Code weekly usage limits by 50% for the promo window, and that boost is extended through the same date. On July 20 your Claude Code limits return to normal.

What does not end: everything else in your plan. Sonnet 5, the Opus 4.x family, your weekly limits (minus the boost), Projects, Skills, all unchanged. Fable 5 also stays fully available on the API, where it is billed separately at standard rates and was never part of this promotion.

One thing Anthropic has not published: a firm date for Fable 5 returning to subscriptions. The original announcement framed this as a compute-capacity problem, and reporting around the extensions says Anthropic intends to restore included access once it has the capacity. Treat that as intent, not a promise with a date.

What happens if I do nothing?

The do-nothing default is fine for most people, which is worth saying plainly since the countdown coverage implies otherwise.

On July 20, Fable 5 stops being usable through your plan's included limits. Anthropic's guidance for the earlier July 7 cutoff was explicit: without usage credits enabled, you will not have access to Fable 5. You switch to another model and keep working within your remaining plan limits. Nothing is charged. Nothing breaks. Your conversations are still there, and Claude will continue them on whatever model you pick.

The model you land on matters, and this is the underrated part: Sonnet 5 supports a 1M token context window on all paid plans when chatting with Claude, per Anthropic's context window documentation. Same window size as Fable 5. For long documents, big codebases in context, and extended conversations, the thing most people actually feel day to day, the free fallback does not lose ground.

How do I enable usage credits and set a spend cap?

If you decide you do want metered Fable 5 access, the setup takes about three minutes, per Anthropic's usage credits article:

1. Open claude.ai, go to Settings, then Usage.
2. Find the Usage credits section and click Enable.
3. Add a payment method if you have not already.
4. Click Adjust limit and set a monthly spending cap. Do this before anything else. The alternative is "Set to unlimited," which I would not leave on for a model that bills output at $50 per million tokens.
5. Click Add funds and buy your starting balance.

Two details that catch people. First, if you subscribed through the iOS or Android app, you can only enable and buy credits on the web version of Claude, not in the mobile app. Second, auto-reload is optional: it tops up your balance when it drops below a threshold you set. Convenient, but pair it with the monthly cap so a runaway agent session cannot quietly compound. Usage credits work on Pro, Max 5x, and Max 20x; Team and Enterprise admins manage them through a separate org-level flow.

How much does Fable 5 actually cost per day of real use?

Here is honest math instead of a scary sticker price. The rates are $10 per million input tokens and $50 per million output tokens, and two forces pull in opposite directions.

Pulling costs down: caching. Long Claude sessions are cache-heavy, and cache reads bill at roughly 0.1x the input rate, so most of a long conversation's context costs about $1 per million tokens instead of $10. The full mechanics are in my prompt caching guide.

Pulling costs up: output. There is no cache discount on output, and Fable 5's thinking is always on and billed as output tokens. The $50 side of the meter is where the money actually goes.

With those assumptions labeled, three realistic tiers:

For calibration on the top tier: the code-tester agent I run for the Cheat Sheet processes about 40M input and 8M output tokens per week. At Fable 5 rates that workload would cost roughly $800 a week, which is exactly why it runs on Opus 4.8 instead. Metered Fable 5 is a scalpel, not a daily driver.

Should I pay for Fable 5 or stay on Sonnet 5?

For most people, stay on Sonnet 5. That is not a budget consolation. Sonnet 5 has the same 1M context window in chat, it is genuinely strong at coding and agentic work, and it is included in the plan you already pay for. If your work is writing, analysis, research, standard coding sessions, and document work, you will rarely feel the difference, and never at a price of $50 per million output tokens.

Pay for Fable 5 credits when your work has both of these properties: the task is hard enough that model quality is the bottleneck (long-horizon agent runs, gnarly multi-file refactors, reasoning problems Sonnet visibly fumbles), and the output of that task is worth real money to you. A consultant whose deliverable bills at $2,000 should not blink at a $15 model spend to get it right. A student summarizing PDFs should not spend a cent.

The honest middle path: enable credits with a low cap, $10 or $20 a month, and treat Fable 5 as an escalation button. Run everything on Sonnet 5, switch models only when Sonnet gets stuck. That converts the deadline from a subscription decision into a per-task decision, which is what metered pricing is actually good at.

How do I cut what Fable 5 costs me?

If you do buy credits, token discipline pays off immediately because every wasted token now has a visible price. The full playbook is in my post on cutting AI token costs, but the three moves with the biggest effect on a metered Fable 5 bill: keep a stable prompt prefix so caching does its 90% work, stop pasting whole documents when a section answers the question, and write prompts that get the answer right on the first pass, because a re-run doubles both the input and the expensive output. If you want ready-made agent setups that already do this, the 20 free agents at clskillshub.com/agents are email-delivered and free, and the dev agents in that pack carry explicit token-budget instructions that were written for exactly this metered world.

What to do before July 19: the 15-minute checklist

1. Export or pin anything important from your Fable 5 conversations (2 min). Chats survive the cutoff, but if a thread must continue on Fable 5 specifically, know that it will need credits after Sunday.
2. Decide credits yes or no (3 min). Use the decision framework above. Default answer is no.
3. If yes, enable now and set the cap (3 min). Settings, then Usage. Cap first, funds second.
4. Use the boosted limits while they last (rest of the window). Claude Code limits are 50% higher through Sunday night. If you have a big refactor or a batch of long-horizon tasks queued, this weekend is the cheapest they will ever be.
5. Note the sibling deadline (1 min). Five days after this one, on July 24, Anthropic removes fast mode on Opus 4.7. If you run API workloads, that one errors instead of gracefully degrading. Two deadlines, one week, handle both in the same sitting.

One more thing worth doing before the meter starts: make your prompts earn their tokens. The Cheat Sheet is 120 tested prompt patterns with the exact structure I use for each, built to get the right answer on the first pass and to cache well. In a world where a sloppy Fable 5 re-run costs real money, prompts that land first time pay for themselves fast. $15, everything included.