Key Takeaways

- Claude subscribers will pay $10 per million input tokens and $50 per million output tokens for Fable 5 access starting July 12
- This is the first time a major AI lab has gated a consumer model behind usage-based billing on top of subscriptions
- Anthropic cites compute capacity constraints and says it aims to return Fable 5 to flat subscriptions when possible
Anthropic is breaking from the AI industry's subscription-only consumer model. Starting July 12, subscribers to Claude's $20, $100, and $200 monthly plans will pay additional usage-based fees to access Claude Fable 5, the consumer version of the company's Mythos 5 model. This appears to be the first time a frontier AI lab has charged consumers per-token fees on top of a subscription.

The rates mirror Anthropic's API pricing: $10 per million input tokens and $50 per million output tokens. A subscriber who sends and receives a million tokens each would owe $60 on top of their base subscription. That's $80 total for the month on the lowest tier, roughly the cost of five months of Amazon Prime.
How much will Fable 5 actually cost heavy users?
A million tokens sounds like a lot. It translates to roughly 750,000 words, longer than the entire Lord of the Rings trilogy. But modern AI models, particularly those with extended reasoning capabilities, can burn through tokens quickly. Hidden chain-of-thought processes inflate output counts. Power users on API plans already rack up thousands of dollars monthly.
The move follows a promotional period where Anthropic offered Fable 5 to subscribers at no extra charge. In its June 7 announcement, the company warned that demand would be "very high, and difficult to predict." Interest spiked further after the US government briefly banned the model for foreign nationals before approving general release on July 1.
Why Anthropic is moving away from flat subscriptions
Anthropic spokesperson Reem Ateyeh told WIRED the company aims to return Fable 5 to flat subscription plans "when sufficient capacity allows" and intends to do so "as quickly as we can." The phrasing points to infrastructure constraints. Despite multibillion-dollar data center deals with Amazon, Google, and SpaceX, Anthropic still lacks the compute capacity it wants.
But the shift also reflects broader economics. AI agents like Claude Code and OpenAI's Codex consume far more compute than traditional chatbots. Nick Turley, OpenAI's former head of ChatGPT who now runs enterprise products there, put it bluntly in a recent podcast: "It's possible that, in the current era, having an unlimited AI plan is like having an unlimited electricity plan. It just doesn't make sense."
Anthropic isn't alone in rethinking pricing. AI coding tools like Cursor overhauled their unlimited subscriptions last year in favor of usage-based models. Anthropic recently started billing large enterprise customers based on actual employee usage rather than flat fees. Some observers speculate these changes are meant to clean up the books ahead of a planned IPO.
A test of consumer willingness to pay
This pricing change doubles as a market experiment. Anthropic has focused primarily on enterprise sales, but Claude is gaining ground with consumers. According to Sensor Tower, Claude reached 245 million unique visitors in May, more than double its February count. That still trails ChatGPT's 1.11 billion and Google Gemini's 662 million, but the growth rate is notable.
The timing is awkward. Some tech leaders have criticized Anthropic for charging premium prices while training on publicly available data. The company is betting it can position itself as the "Apple of the AI Era," with a segment of users willing to pay more for perceived quality and safety. Whether that bet pays off at the consumer level remains an open question.
What this means for AI product teams
For teams building on Claude's API, the economics haven't changed. You were already paying per token. But if you've been prototyping with the consumer product to avoid API complexity, that shortcut just got expensive.
The bigger signal is directional. If usage-based billing becomes standard for consumer AI products, it will reshape how users interact with AI tools. Workflows built around unlimited access will need to account for cost. Product teams integrating third-party AI will need to model variable expenses more carefully.
Logicity's Take
Anthropic's move isn't surprising given compute economics, but the execution matters. Layering per-token fees on top of subscriptions without a usage dashboard or spending caps creates friction. Compare this to OpenAI's ChatGPT Plus at $20 flat, or Google's Gemini Advanced at $20, both of which still offer their flagship models without metered billing. Anthropic is betting on model quality justifying premium pricing. For AI product teams building agents or high-throughput applications, this validates what API users already knew: inference costs are the new infrastructure line item.
Frequently Asked Questions
When does Claude Fable 5 usage-based pricing start?
July 12, 2025 at 11:59 PM Pacific Time. After that date, all Claude subscribers accessing Fable 5 will pay per-token fees on top of their subscription.
How much does Claude Fable 5 cost per token?
$10 per million input tokens and $50 per million output tokens, matching Anthropic's API rates.
Will Anthropic return Fable 5 to flat subscription pricing?
Anthropic says it aims to do so "when sufficient capacity allows" but has not provided a timeline.
How does Claude Fable 5 pricing compare to ChatGPT?
ChatGPT Plus remains $20/month with access to GPT-4o and other models without per-token charges for most uses. Claude's new model adds variable costs on top of the base subscription.
Why is Anthropic adding usage-based fees?
The company cites compute capacity constraints. More capable AI models, especially those with reasoning features, consume significantly more resources than traditional chatbots.
Need Help Implementing This?
Logicity helps AI product teams navigate API pricing, model selection, and cost optimization. If you're evaluating Claude, GPT-4, or Gemini for production workloads, reach out for a free consultation.
Source: Feed: Artificial Intelligence Latest / Maxwell Zeff
Huma Shazia
Senior AI & Tech Writer
Produced with AI assistance and reviewed by the Logicity editorial team. Learn more in our Editorial Policy.
Related Articles
Browse all
AI Search Trust Problem: Why 85% of Users Doubt Results
New research reveals a massive gap between AI search adoption and user trust. Two-thirds of Americans use AI search tools, but only 15% trust the results. For businesses relying on AI-powered discovery, this trust deficit represents both a risk and an opportunity.

INSIDER REVEAL: How the American Enterprise Institute Uncovered the AI Productivity Boom
The American Enterprise Institute has been searching for signs of an AI-driven productivity boom. According to McKinsey, AI can increase productivity by up to 40%. We dive into the details of this emerging trend and what it means for businesses.

Will AI Ethics Regulation Become the New Industry Standard?
The Vatican has emphasized the need for AI ethics regulation in a recent statement, sparking a global conversation about responsible AI development. We explore the implications of this call to action and what it means for businesses and individuals alike. As AI continues to shape our world, we must consider the ethical implications of its development and deployment.



