Claude Fable 5 Access After the June 22 Deadline: Where Things Stand

Official Sources#

Last verified: July 29, 2026

When Anthropic launched Claude Fable 5 on June 9, it came with a clock attached: Fable 5 was included on Pro, Max, Team, and seat-based Enterprise plans at no extra cost, but only through June 22. That window is now closed. If you're reading this after the fact, here's what actually happened when the free period ended, what access looks like now, and where to go if you're still deciding whether Fable 5 is worth using at all.

Anthropic's stated reason for the time limit was capacity: "We expect demand for Fable 5 to be very high, and difficult to predict." They said they'd extend the included window if capacity allowed, and that the goal was to restore Fable 5 as a standard part of subscriptions "as quickly as we can." Neither promise came with a date, and neither has materialized as of this update.

Here's what changed, what stayed the same, and how to plan going forward.

The timeline#

• June 9 to 22: Fable 5 included free on Pro, Max, Team, and seat-based Enterprise plans.
• June 23 onward: Fable 5 came off those plans. Continued access now runs through usage credits, layered on top of the existing subscription rate limits rather than replacing them.
• Unchanged throughout: API access ($10/$50 per MTok), consumption-based Enterprise, and cloud platforms (Bedrock, Vertex, Foundry). The June 22 cliff only ever affected subscription plans.
• Also unchanged: everything else on your plan. Opus 4.8, Sonnet 4.6, and Haiku 4.5 stayed exactly where they were.

What credit access actually costs now#

Anthropic didn't publish a separate subscriber credit rate card ahead of the cutoff, and the practical answer for most people has settled on the API price as the working anchor: $10 per million input tokens, $50 per million output tokens, exactly double Opus 4.8. If you're budgeting for Fable 5 usage past your plan's included allowance, use those numbers and see how Fable's usage windows and multipliers work in practice before you commit to a number. For any workload you plan to run in production rather than occasionally, build a real cost model first - the effort-level spread alone can swing a task's cost by 7x.

What the burn actually looked like#

Fable 5 is hungry. Real numbers from the first 48 hours after launch, before anyone had learned to budget for it:

• Simon Willison burned $110 worth of tokens in a single day inside a Max subscription, including one session that consumed 78 million tokens (about $99 at API rates).
• One user reported hitting the $200/month Max plan's limits in under 30 minutes.
• Heavy users report individual tasks routinely consuming 500K to 1M tokens.
• Anthropic had to reset rate limits across products on launch day because of demand.

Two reasons it adds up fast. Single requests at high effort can run many minutes and think for a long time before answering, and the model is built for long autonomous sessions, which means multi-hour runs that consume tokens the whole time.

One counterweight: a pre-launch tester reported Fable 5 used roughly half the tokens of Opus 4.8 to complete the same agentic work. The unit price is 2x, but the effective cost on long tasks lands closer to Opus than the sticker price suggests. Short interactive work doesn't get that efficiency win, so chat-style usage really does cost about double.

What to do now, by user type#

You're on Pro or Max and never tested it during the free window: you can still find out whether Fable 5 beats Opus 4.8 on your work, it just costs credits now instead of nothing. Test it on your hardest problems, not routine ones - Anthropic's own guidance is that easy tasks undersell the model, and early consensus was that output on routine work felt identical to Opus 4.8. If your day-to-day is routine, stay on Opus 4.8 and skip the credit spend entirely.

You're on Max and Fable 5 is your daily driver: the "plan with one model, execute with the other" split that emerged during the free window still holds. Run planning, spec work, and cheap iteration on Opus 4.8 or Sonnet 4.6, and reserve Fable 5 credits for the genuinely hard, long-horizon jobs where its lead over Opus actually shows up.

You manage a team plan: if you ran a usage audit during the June 9-22 window, that data is still your best procurement signal - use it to decide who actually needs standing credit access versus API keys for occasional use. If you skipped the audit, run one now on a smaller scale before committing budget across the whole team.

You're API-only: nothing about your setup changed on June 22 or since. The decision is still the standard one: does Fable 5 justify 2x Opus pricing on your workload? Run both on identical tasks and measure cost per completed task, not cost per token. Effort levels are the biggest lever you have on that number.

Three ways to control cost now that the meter's running#

1. Use effort levels. Fable 5 has five (low through max) with a wide cost spread between them. Anthropic recommends starting at high, not xhigh. Independent testing found the same task varied from about 10 cents to about 72 cents depending on effort level - see the full effort-level breakdown for what each tier actually changes.
2. Match the model to the task. Fable 5's lead over Opus 4.8 grows with task length and complexity. Inverting that: the shorter and more routine the task, the less you get for the 2x price. Default to Opus 4.8 or Sonnet 4.6; escalate to Fable 5 when a task is genuinely hard.
3. Model the cost before you scale it. A single test run doesn't tell you what a week of production usage costs. Build the spreadsheet first - it's a bigger cost lever than any single prompt tweak.

If your workload involves feeding Fable 5 large amounts of source material - full repos, long log files, multi-document research - the model's default 1M-token context window changes what's practical, but it's easy to overspend by stuffing the window when retrieval would be cheaper. See what actually works in practice before you architect around it.

The bigger picture#

It's fair to read this launch as a preview of how frontier-model access works now: the newest tier costs real money, and subscriptions cover the tier one notch down. The top Hacker News thread on the launch called the two-week window the "free sample method," and the comparison held up. Anthropic hasn't restored free inclusion as of this update, and hasn't given a date for when or whether it will.

The practical version of that story is still less dramatic than the framing suggests. Opus 4.8 remains on every plan, remains excellent, and is still what Fable 5 falls back to when its own safety classifiers fire. For most work, the model already on your plan is the right tool. Fable 5 credit spend is best reserved for the slice of your work where the gap actually matters.

FAQ#

Is Claude Fable 5 free?#

No, not anymore. Per Anthropic's launch announcement, Fable 5 was included on Pro, Max, Team, and seat-based Enterprise plans at no extra cost only through June 22, 2026. Since June 23, continued use on those plans draws from usage credits. There was never free-tier access, and API usage has been billed from day one at $10 per million input tokens and $50 per million output tokens. See how the usage windows and credit multipliers work for the current mechanics.

Did the June 22 deadline affect Claude API access to Fable 5?#

No. The June 22 cutoff only ever affected subscription plans. API access at $10/$50 per million tokens, consumption-based Enterprise, and the cloud platforms (Bedrock, Vertex, Foundry) were unchanged throughout. Opus 4.8, Sonnet 4.6, and Haiku 4.5 also stayed exactly where they were on every plan.

How can I keep Fable 5 costs under control now that it uses credits?#

Three levers matter most. Use effort levels - Anthropic recommends starting at high, not xhigh, and independent testing found the same task varied from about 10 cents to about 72 cents depending on effort. Match the model to the task by defaulting to Opus 4.8 or Sonnet 4.6 and escalating to Fable 5 only when a task is genuinely hard. And model the cost before you scale it, since a single test run does not tell you what a week of production usage costs.

Sources: Anthropic's launch announcement, pricing docs, Simon Willison's first-day writeup, and the Hacker News launch thread.

Continue Reading#

• How Claude's Usage Limits Actually Work With Fable 5 - windows, multipliers, and burn rates now that credits are live
• Fable 5 Effort Levels Explained - what low through xhigh actually change and cost
• How to Model Fable 5 Costs Before They Blow Up Your Budget - build a real cost model before you commit
• Fable 5 with 1M Context: What Actually Works in Practice - the workflows worth the token spend
• Migrating to Claude Fable 5: The Practical Guide - breaking changes and the prompt audit to run first