Anthropic extended free access to Claude Fable 5 — its most capable publicly available model — through Sunday, July 12, just hours before the original deadline would have moved the model to per-token credit billing. The extension applies automatically to all Pro, Max, Team, and select Enterprise plans, giving subscribers who haven't yet used the model, or who burned through their allowance anticipating the July 7 cutoff, a second chance to put the Mythos-class model through its paces at no additional cost.

The decision matters if you work with long documents, complex multi-stage coding projects, or tasks that require the model to work autonomously for extended periods — those are exactly the workloads Fable 5 was built for, and they are the workloads that will cost the most once the window closes Monday morning.

Why Anthropic Extended, and What the Extension Covers

The extension shifts the free-access deadline by five days, from July 7 to July 12. Nothing else changed. The 50 percent weekly usage-limit cap remains in place, the plan coverage remains the same, and the credit pricing that takes effect after the window closes — $10 per million input tokens and $50 per million output tokens — remains unchanged. That rate is the highest Anthropic has published for any generally available model and is exactly double the rate for Claude Opus 4.8.

Anthropic did not issue a formal blog post. The announcement came through the company's official X account on Tuesday, July 7. Subscribers whose Claude interface displayed a small popup notification about the extension have confirmed the update is live.

The company has stated the shift to credit billing after July 12 is a temporary capacity measure, not a permanent pricing decision. A Claude Code lead engineer confirmed publicly that Anthropic aims to return Fable 5 to standard subscription plans once compute capacity allows.

What Makes Fable 5 Different From Opus 4.8 — and Why the Window Is Worth Using

Fable 5 belongs to Anthropic's Mythos class, a model tier introduced in 2026 that sits above the Opus class in capability. The two share the same underlying model architecture; what separates them are the safety classifiers Anthropic applied specifically to Fable 5 to make it suitable for general use.

The practical difference for subscribers shows up in three areas. First, Fable 5 supports a 1-million-token context window, meaning it can hold an entire data room, codebase, or research archive in a single session. Second, it is designed for multi-day autonomous operation — tasks that Anthropic describes as "days-long, complex, and asynchronous" that previous models couldn't sustain. Third, it includes extended thinking, where the model reasons through a problem internally before responding, returning a summarized version of that reasoning rather than the raw chain of thought.

None of those capabilities are available at Opus 4.8 or Sonnet 5 pricing. Claude Sonnet 5, which launched June 30 as the new default for Free and Pro plans, runs at $2 per million input tokens — one-fifth the Fable 5 rate — and handles most agentic coding well. For workloads that don't specifically require Fable 5's long-horizon coherence or its 1-million-token context window, Sonnet 5 is the cost-efficient alternative once the free window closes.

One technical behavior subscribers using Fable 5 through Claude.ai, Claude Code, or Claude Cowork should know: the model includes safety classifiers that can automatically decline certain cybersecurity-adjacent requests and reroute them to Opus 4.8. Those rerouted sessions are billed at Opus 4.8 rates, not Fable 5 rates. On the raw API and Amazon Bedrock, this fallback is not automatic — developers must handle refusals in their integration code.

A Turbulent Six Weeks Made This Window Shorter Than Planned

The July 12 deadline is the end of the second free-access window for Fable 5, not the first. The model launched June 9 with a planned two-week subscription-included period running through June 22, with no usage-percentage cap. Three days into that window — before most subscribers had meaningfully tested the model — the US Commerce Department issued an export-control directive requiring Anthropic to restrict both Fable 5 and Mythos 5 to US nationals only. Because Anthropic had no mechanism to verify citizenship at consumer scale in real time, the company disabled both models for all users globally within about 90 minutes of receiving the order.

The 19-day suspension ended June 30, when Commerce Secretary Howard Lutnick withdrew the directive. Anthropic redeployed Fable 5 on July 1 with a compressed six-day window — July 1 through July 7 — at a reduced 50 percent weekly usage cap. Tuesday's extension adds five more days to that compressed window, bringing the total effective subscription-included time to roughly eleven days, still less than the two weeks originally promised at launch.

What Happens After July 12: Enable Credits Before the Window Closes

Starting Monday, July 13, all Fable 5 usage on subscription plans requires a funded usage-credit balance. Credits are a separate, prepaid billing layer on top of the standard subscription. To enable them: go to Settings, then Usage, enable the feature, and add funds. A daily redemption cap of $2,000 applies, and auto-reload is available.

Two details matter operationally. First, if credits are not enabled before the free window closes, Fable 5 access simply stops with no grace period and no automatic fallback to another model — the model becomes inaccessible until credits are funded. Second, the credit rate for Fable 5 agentic work accumulates differently than a flat subscription. One developer publicly reported burning through a $100 daily credit allocation in nine minutes during an autonomous coding loop. A single long agentic session generating substantial output tokens can represent a meaningful cost at $50 per million output tokens.

Prompt caching reduces input token costs by 90 percent on cached content — from $10 to $1 per million tokens — which meaningfully lowers the cost of sessions that reuse a large system prompt or document set. The Batch API offers a 50 percent discount for non-urgent, asynchronous work. For tasks that do not require Fable 5's specific capabilities, Sonnet 5 at its introductory rate remains the best cost-performance option through at least August 31, 2026.

Fable 5 and Standard Enterprise Seats

Not all Enterprise subscribers share the same access. Standard Enterprise seats have never included Fable 5 in their subscription allowance. Access for those users has always required usage credits to be enabled at the organizational level. Premium Enterprise seats have had the same July 7 (now July 12) free-window terms as Pro, Max, and Team plans.

If you are on a standard Enterprise seat and attempting to use Fable 5, the first step is confirming with your organization's admin whether usage credits are enabled — not waiting for the July 12 extension to apply, because it does not apply to that plan tier.

Frequently Asked Questions

What should I do right now to take advantage of the July 12 extension?

Open Claude and confirm which model is selected — Fable 5 must be manually selected in most interfaces rather than set as the default. Use the remaining free window for your most demanding tasks: large-context document analysis, complex multi-step coding projects, or long-form research synthesis that requires the 1-million-token context window. Tasks that fit within Opus 4.8 or Sonnet 5's context limits and don't require multi-day autonomous operation are better saved for those models after July 12, when per-token billing begins.

What exactly happens to Fable 5 access after July 12?

After July 12 at 11:59 PM PT, Fable 5 becomes available only through usage credits at $10 per million input tokens and $50 per million output tokens. If you have not enabled and funded usage credits before the window closes, access to Fable 5 stops immediately with no grace period. Anthropic has stated this is a temporary capacity measure and that the company plans to restore Fable 5 to standard subscription plans when compute capacity allows, though no timeline has been given for that restoration.

How fast can usage credits run out in real-world use?

At the published credit rate, usage accumulates quickly for agentic work. A developer running Fable 5 inside an autonomous coding loop publicly reported burning through $100 in nine minutes. Casual chat sessions consume far less, but any workflow that generates long outputs or operates Fable 5 autonomously across many steps should be budgeted carefully. Setting a monthly spending cap in Settings under Usage limits exposure to unexpected charges.

Is Claude Sonnet 5 a reasonable alternative once the free window closes?

For most tasks, yes. Sonnet 5 launched June 30 at $2 per million input tokens and $10 per million output tokens — one-fifth the Fable 5 rate — and runs with near-Opus 4.8 performance on agentic coding and standard knowledge work. It does not support the 1-million-token context window or the multi-day autonomous session capability that defines Fable 5's advantages. For workloads that fit within those constraints, Sonnet 5 is the most cost-efficient option available and remains on introductory pricing through August 31, 2026.