← Back to feed

Effective today, every Claude subscriber — Pro, Max, Team, and Enterprise — who wants to use Fable 5 needs a funded usage-credit balance. The grace period that began on July 1 when the model returned from its 19-day export-control shutdown has ended. From here, Fable 5 draws exclusively from metered credits at $10 per million input tokens and $50 per million output tokens.

That pricing makes Fable 5 the most expensive generally available model Anthropic has ever shipped — double the rate of Claude Opus 4.8 ($5/$25) and roughly on par with GPT-5.5 High ($5/$30 output) for output-heavy workloads, while costing significantly more on input.

How the Billing Works

For Pro, Max, and Team subscribers, Fable 5 had been included for up to 50% of weekly usage limits since its July 1 return. That inclusion is gone. All Fable 5 sessions now consume from a separate usage-credit balance that must be manually enabled and funded before the first session. If credits run out mid-session — in the middle of an agentic loop, mid-file, mid-analysis — access stops. There is no automatic fallback to Opus 4.8 or any other model.

For Enterprise seats, the transition is similarly clean: standard Enterprise plans never had included Fable 5 access, and premium Enterprise seats lose their grace allowance today.

The Six Weeks That Got Here

Fable 5 launched June 9 as Anthropic’s first publicly available Mythos-class model — a tier above Opus for ambitious, long-running agentic work. It was initially included free on Pro, Max, Team, and Enterprise through June 22, at which point it moved to usage-credit billing for the first time.

Three days after launch, the US Commerce Department issued an export-control directive requiring Anthropic to suspend both Fable 5 and Mythos 5 for foreign nationals. Lacking real-time nationality verification, Anthropic shut both models globally within 90 minutes. The 19-day suspension wiped out the original June 22 free window for nearly every subscriber. When the models came back on July 1 after Anthropic satisfied a set of government technical conditions, a new six-day included window was established as compensation. That window closed last night.

The Agentic Cost Math

The $50/M output rate is where the real impact lands. Fable 5 is optimised for long, autonomous runs. A single agentic coding session that generates 100,000 output tokens — not uncommon for a multi-file refactor — costs $5 in Fable 5 output alone, versus $2.50 for Opus 4.8 and $3 for GPT-5.5 standard. Multiply across a development team running several agents per day and the monthly cost differential becomes meaningful.

The benchmark justification for the premium: 95.0% SWE-bench Verified (vs 88.6% for Opus 4.8), 80.3% SWE-bench Pro, Terminal-Bench 2.1 at 84.3%. For teams with workloads sensitive to that capability gap, the math may still work. For teams that were running Fable 5 because it was included and Opus 4.8 was technically sufficient, today’s transition is the natural forcing function to find out.

Pricing at a Glance

Model	Input/M	Output/M
Claude Fable 5	$10	$50
Claude Opus 4.8	$5	$25
GPT-5.5	$5	$30
Claude Sonnet 5	$2 (intro)	$10 (intro)

For organisations that need frontier capability without the Fable 5 price point, Claude Sonnet 5 is now the default across Free and Pro accounts at introductory rates of $2/$10 through August 31.