Anthropic expects to earn 18 billion dollars this year. Some analysts estimate that they have already completed their targets ahead of schedule. But clearly, they are not satisfied.

Early this morning, a developer named George Pu mentioned his discovery on X: "Anthropic (the parent company of Claude) has just removed Claude Code from the Pro plan. Pro users wanting to use this feature must now upgrade to the Max plan, with a minimum monthly fee of $100."

This means that to maintain the same permissions, the price has increased fivefold. There are no blogs, no emails, no changelogs. Anthropic conducted a very quiet page update, attempting to raise prices.

This tweet quickly spread. Screenshots of the pricing page began circulating on Reddit, Hacker News, and developer groups. The community repeatedly confirmed on their screens that there was a cross under the Claude Code line below the Pro column, with only Max 5x ($100/month) and Max 20x ($200/month) still checked. Meanwhile, Anthropic's support document previously stated "Using Claude Code with your Pro or Max plan," now it has changed to "Using Claude Code with your Max plan."

The community was naturally very dissatisfied, and as public pressure built up to a certain extent, Anthropic’s growth lead Amol Avasare felt compelled to respond with a post on X. He stated that this was simply a small-scale test targeting about 2% of new Pro users and that existing Pro and Max subscribers would not be affected.

He later posted a longer explanation in the same thread, meaning that: When Max was launched a year ago, it did not include Claude Code, Cowork did not exist either, and long-running agents were not part of the daily workflow. The design goal of Max was only for heavy chat users.

However, after the release of Opus 4, the usage of Claude Code exploded. Asynchronous agents became daily tools for developers. "Subscriber per capita usage has significantly increased, and our current pricing model was not designed for this kind of usage."

He ultimately assured that if the final plan affects existing subscribers, they would provide ample prior notice, "the news will come from us, not from screenshots on X or Reddit."

In recent months, power supply has been continuously tightened

Avasare's explanation sounds reasonable, but when pieced together, it points to a more fundamental issue: Anthropic may be experiencing severe power supply pressure.

Each coding session of Claude Code consumes a token amount far exceeding that of regular conversations, sometimes differing by an order of magnitude. As users flocked after the release of Opus 4, when agents began running asynchronous tasks for several hours, and when Cowork drew more people into high-usage scenarios, the marginal cost of the $20/month price began to spiral out of control.

From a business logic perspective, this is not a minor annoyance of "users using too much," but rather a fundamental misalignment of the pricing model after the product capabilities evolved rapidly. Initially, Max was designed for heavy chatting; now Max has to support continuously running programming agents. The computational cost of these two aspects is completely of a different order.

Avasare used a phrase: "engagement per subscriber is way up." Translated into business language, it means: the resources users are consuming now for the same amount of money have increased significantly.

This is not the first time. Extending the timeline, Anthropic has made quite a few "power supply tightening" moves in recent months.

In August 2025, Anthropic announced weekly usage limits for Pro and Max users, effective August 28. The official reason was that some users violated the usage terms by sharing and reselling accounts to bypass restrictions. The statement claimed that "this is expected to affect less than 5% of subscribers." The weekly limit marks the first time a weekly hard cap has been set for high-paying users.

From late August to early September of the same year, things became more complicated. Numerous developers reported on Reddit and X that the code generation quality of Claude had sharply declined, as the model began to ignore its plans and randomly produce gibberish, making the task execution of Claude Code unreliable. Anthropic eventually admitted that this was a technical bug that emerged during the reasoning stack upgrade, causing Claude Opus 4.1 to continue degrading in quality for about 56.5 hours and released a formal postmortem report.

This bug incident soon overlapped with a larger controversy.

In early March 2026, a senior AI director at AMD, Stella Laurenzo, released a detailed analysis on GitHub, based on 6852 Claude Code session files, 17871 thinking blocks, and over 230,000 tool calls, concluding that:

Claude Code had dumbed down. This analysis resonated strongly in the developer community.

Boris Cherny, the person in charge of Claude Code, responded thereafter, denying the core conclusion of "the model has worsened," explaining that Opus 4.6 switched to "adaptive thinking" mode on February 9, and on March 3, adjusted the default effort level to medium (85 tier), which the officials considered to be "the best balance between intelligence, delay, and cost." However, he also acknowledged that about 7% of users would touch session limits they previously wouldn’t, and this adjustment primarily targeted Pro users during peak periods, while Team and Enterprise were not affected.

From March 13 to 27, 2026, Anthropic launched a "double usage during non-peak hours" promotion: on weekdays from 2 PM to the next day 8 AM (Eastern Time), and all weekend, usage limits would automatically double, covering all plans: Free, Pro, Max, and Team.

On the surface, it seemed like a welfare activity. But in light of the background, the reason for doubling during non-peak hours is that peak hours are currently experiencing tight power supply. Offering double limits for idle capacity does not affect congested periods while creating motivation for upgrades to higher-tier plans.

Then on April 4, Anthropic announced a ban on third-party agent frameworks running through Pro and Max subscription quotas. Users who wish to continue invoking Claude through external frameworks like OpenClaw must pay additionally per use.

And then there is this matter of Claude Code disappearing from the Pro pricing page.

A is certainly not the last to do this

Returning to Avasare's tweet. He mentioned that "Claude Code disappearing from the Pro pricing page is only a small-scale test targeting about 2% of new Pro users," but this statement sparked a lot of skepticism in the community.

The most direct question came from developer Simon Willison: "Everyone I know has seen the new pricing page; the Internet Archive has also stored a screenshot." If it truly only affects 2% of users, why was the entire public pricing grid changed? Why was the support document updated across the board?

More intriguingly: when asked by reporters why the pricing page and support document underwent extensive changes if the test only targeted 2% of users, he did not respond. Similarly, Anthropic's company spokesperson did not answer this follow-up question.

Another contradiction was noted: Claude Cowork was not removed from the Pro plan, but from a product functionality perspective, Cowork is essentially just a rebranded Claude Code. In the words of one developer, "Cowork is Claude Code dressed in a less intimidating disguise."

The development was that after several hours of public opinion fermentation, Anthropic quietly restored the website and support documents to their original state. No explanations, no announcements. Just like the change itself, it was quietly completed.

If we look at these actions together, we will find that Anthropic is now acting cautiously, as everyone can see:

Anthropic is testing higher power usage boundaries for subscription users in a tentative manner, step by step. Doubling usage on weekends is guiding demand to shift to off-peak times; lowering the default effort level saves power during peak hours; banning third-party agent frameworks is plugging loopholes in quota arbitrage; and Claude Code disappearing from the Pro page is testing users' reactions to a more aggressive repricing scheme.

Amol himself admitted: "Our current plan was not designed for this usage." This is a rare admission. It indicates that Anthropic has realized, to some extent, that the current pricing structure cannot sustainably support the computational consumption brought by Claude Code.

What changes will happen next, we do not know.

But perhaps like Ed Zitron wrote in his article: "I do not believe this will be the last time Anthropic does this."

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。