Anthropic is changing the weekly usage limits for Claude Code, its in-house AI programming tool, starting September 14. From this date, a limit that is 25 percent higher than the original base limit will apply permanently – compared to the currently running transitional regulation with a 50 percent surcharge, this means a decrease of 17 percent. Affected are Pro, Max, Team, and seat-based Enterprise subscriptions.
Transitional phase with increased limit ends after four months
Since May 13, Anthropic has been operating Claude Code with a temporary surcharge of 50 percent on the published standard limits for Pro, Max, Team, and seat-based Enterprise customers. The company has extended this transitional regulation several times, most recently on August 19 until September 13, citing noticeable capacity bottlenecks in its own computing infrastructure due to the rapidly growing demand for the programming tool since its market launch. Users were able to submit significantly more requests, generate code, and run agents during this period than the original limits would have allowed.
As of September 14, this phase will now end definitively. As Anthropic announced via the official account Claude Devs on X, the standard limits will be permanently raised by 25 percent compared to the base limit – a permanent improvement that, however, is below the level users recently grew accustomed to. In a second post immediately following, the company confirmed that this corresponds to a reduction of 17 percent compared to the current state.
Calculation example shows the concrete impact
The difference can be illustrated with a simple example that several trade media, such as Bleeping Computer, adopted from the announcement. If the original weekly limit was around 100 units, users currently have 150 units available thanks to the 50 percent surcharge. Starting September 14, this value will drop to 125 units: an increase of 25 percent compared to the base, but a decrease of 17 percent compared to the current state. The figure of 17 percent comes directly from Anthropic and is not independently verified.
Anthropic does not publish absolute numbers for the actual weekly limits; how many requests or lines of code a limit encompasses in detail also depends on conversation length, chosen model, tool usage, and set effort level. Anthropic points out that limits therefore cannot be translated into a fixed number of prompts. For heavily used developer teams, the reduction is likely to be noticeable, especially for those who have grown accustomed to the larger leeway in recent months.
Anthropic promises more transparency in usage
Anthropic has been rather reserved in its justification. The post only refers to changes intended to give users the feeling of getting more out of Claude, along with greater overview and control over their own consumption. The company has so far not provided details on this. Anthropic had already taken a first step in this direction in July with the usage dashboard Claude Reflect, which makes usage patterns visible over several months – so far, however, without direct reference to Claude Code itself.
In other areas, Anthropic has also recently adjusted the default settings of Claude Code: since August 14, an automatic classifier replaces the manual approval of risky commands for Pro, Max, and Team subscriptions, after a study showed that people overlook nine out of ten risky commands. Users on the developer forum Hacker News reacted with mixed views to the limit announcement: some criticized the terse wording of the first announcement, while others pointed out that the high surcharge had been communicated as temporary from the beginning.
What will be decisive is how exactly the transparency and control features announced by Anthropic will look and whether they can offset the loss of capacity from the perspective of heavily used teams. It also remains open whether these tools will appear before the September 14 deadline – Anthropic continues to provide no concrete numbers on the actual weekly limits.
