Anthropic announced a change in its business model: the use of Claude Fable 5, its most sophisticated artificial intelligence model intended for end consumers, will now be charged, even for those who already have a monthly subscription to the service.
This modification will take effect on July 12 at 11:59 PM Pacific Time, corresponding to 3:59 AM on the 13th in Brasília time. The company stated that subscribers to the $20, $100, and $200 monthly packages will have to pay extra fees calculated based on the volume of model usage. This is the first notable case of a large AI lab charging consumers this way to access its most advanced feature.
The new charges will follow the pricing structure already used for developers accessing the Anthropic API. The established cost is $10 (equivalent to R$ 51.36) per million tokens sent to Claude Fable 5 and $50 (R$ 256.78) per million tokens generated as a response by the model. As a practical example, a subscriber on the $20 plan who uses one million input tokens and receives one million output tokens will incur an additional cost of $60 (R$ 308.13), totaling $80 (R$ 410.84) that month. It is important to note that one million tokens is equivalent to about 750,000 words, but consumption can be high due to the intensive processing required by recent models that use internal reasoning chains.
Consumption-based billing is already standard among developers using models via APIs. However, in the consumer segment, the market has traditionally favored fixed-value monthly subscriptions, aiming for both revenue generation and demand control. However, this approach is changing. Last year, AI programming tool companies like Cursor abandoned unlimited plans and started charging for the volume used. Anthropic itself had already implemented a similar model for large corporate clients, charging based on employee usage rather than a fixed fee. For WIRED, these changes may be linked to the company's preparations for a future Initial Public Offering (IPO).
Industry experts argue that conventional subscription models have lost relevance with the rise of AI agents, which are capable of performing complex tasks that require much more computational power than common chatbots. Nick Turley, former head of ChatGPT at OpenAI and current head of corporate products at the company, expressed this view in a recent podcast, comparing an unlimited AI plan to an unlimited electricity plan, stating that it simply does not make sense in the current era.
Despite the introduction of usage-based billing, Anthropic stated that it does not intend to permanently discard the traditional subscription model. In communication with WIRED, spokesperson Reem Ateyeh guaranteed that the company plans to reintroduce Claude Fable 5 into subscription plans as soon as computational capacity is sufficient, which will happen as soon as possible. This statement indicates that the company's main current constraint lies in the availability of computing infrastructure. Although Anthropic has signed multi-million dollar agreements with SpaceX, Amazon, and Google to expand its data centers, the existing infrastructure is still below what the company desires, with no forecast for when these limitations will be overcome.
This change comes after a promotional period during which Claude Fable 5 was offered free to subscribers. When the model was launched on June 7, Anthropic expected a 'very high and unpredictable' demand. Since then, interest in the system has grown even more, driven by the temporary ban on model access for foreign citizens by the U.S. government and the subsequent general release on July 1. Regardless of how Anthropic presents the decision, the new pricing policy serves as a test to assess how much consumers are willing to pay for access to what is considered the company's most advanced model. Although it has focused mainly on the corporate market, Anthropic has been expanding its presence in the consumer segment, where OpenAI's ChatGPT and Google's Gemini dominate.
Data from the market intelligence company Sensor Tower, cited in the article, indicates that Claude attracted 245 million unique visitors in May, more than double the number recorded in February. Despite this advance, the service still lags behind the main competitors: ChatGPT registered 1.11 billion monthly unique visitors, and Gemini reached 662 million. Anthropic's strategy aims to position itself as the 'Apple of the AI era,' betting on the continued existence of consumers willing to pay more for a premium product. A source close to Anthropic, speaking anonymously to WIRED, observed that the discussion among users has shifted from 'Do I need the best intelligence for this task?' to 'Am I the type of person who needs intermediate or the best intelligence?'. This source added that many well-paid professionals in the financial, political, and technological sectors already show a willingness to invest in the most advanced model, both professionally and in personal use, viewing Claude as a viable option. However, maintaining this reputation will depend on Anthropic's ability to maintain competitive leadership, given that initial reactions to OpenAI's GPT-5.6 Sun suggest a fierce competition.
However, OpenAI and Google may not follow the same path as Anthropic. These companies often increase advertising use to subsidize their free and low-cost plans. Since Anthropic avoids the use of ads, price increases may be its main short-term revenue expansion route. This change suggests that the so-called 'golden age' of subsidized AI subscriptions for consumers may be coming to an end.
