Fable 5 Exits Included Plans Tonight - $10/$50 Credits Start Tomorrow for Every Subscriber

Fable 5 exits Pro, Max, Team, and select Enterprise plans tonight. Starting July 8, every subscriber who wants to keep using it needs Anthropic usage credits enabled - billed at $10 per million input tokens and $50 per million output tokens, exactly double Claude Opus 4.8's rate and the highest listed price for any generally available Anthropic model. For casual subscribers, that change shows up as an extra line on the next bill. For developers running Fable 5 inside agentic loops, the cost shift is operationally significant enough to audit before it starts accruing overnight.

Getting to this point took six weeks and two billing cliffs. Fable 5 launched June 9, initially free on paid plans through June 22, before Anthropic moved it to usage credits for the first time. Three days into that original run, the US Commerce Department issued an export-control directive requiring Anthropic to suspend the model globally - no mechanism existed to verify citizenship at consumer scale, so both Fable 5 and Mythos 5 went dark for all users worldwide within 90 minutes. Commerce Secretary Howard Lutnick withdrew the order on June 30; access was restored July 1 with a replacement six-day free window. That window closes tonight. The full backstory of the suspension and return is in Claude Fable 5 Returns July 1 After US Lifts Export Controls.

For developers, the $10/$50 headline rate understates the actual cost of running Fable 5 in agent loops. Transformer models are stateless - every accumulated context, plan summary, tool definition, and file chunk must be re-sent as input on every turn. In a multi-hour autonomous coding session, that means a growing context window bills as input on each call while the model's reasoning response - often long, structured, and action-heavy - bills as output at the dominant $50-per-million rate. BleepingComputer's testing found a $100 Max subscription drained in under nine minutes during heavy use. Scrimba's CEO logged 1.3 million tokens in seven minutes, a pace equivalent to about $160 per hour. Three cost-reduction levers are available: prompt caching drops input from $10 to $1 per million on any repeated context such as system prompts or shared codebase scaffolding; the Batch API halves both rates to $5/$25 for non-real-time work; and routing Fable 5 only to tasks that genuinely require its 1-million-token context window keeps spend proportional to benefit.

One complication arrived with the July 1 relaunch. A new safety classifier deployed alongside the model has a documented false-positive rate on routine coding tasks - BridgeMind's testing on a TypeScript debugging benchmark found nine of 12 tasks were rerouted to Opus 4.8 before Fable 5 could complete them. Rerouted sessions bill at Opus rates rather than Fable rates, which lowers cost but also means developers who specifically need Fable 5's long-horizon coherence may not receive it for a significant share of coding tasks. More detail on the classifier behavior is in Fable 5 Returns With 50% Plan Limits and a Safety Classifier That May Block Your Code.

Before midnight, enable usage credits in the Claude Console and set a monthly spending cap - without credits active, Fable 5 access stops with no automatic fallback when the inclusion runs out. Teams that do not specifically need the model should route to Claude Sonnet 5 ($2/$10 per million, introductory pricing through August 31) or open-weight GLM-5.2, which runs at roughly one-sixth Opus 4.8's cost. Anthropic has framed the credits requirement as a capacity-driven measure rather than a pricing change, and a Claude Code lead engineer confirmed publicly the company intends to restore Fable 5 to standard subscriptions once infrastructure allows - without giving a timeline. The irony is that Fable 5 is moving to credits precisely because it works: the demand Anthropic cannot serve at flat subscription prices is the kind of problem most companies would accept.