In recent days, there has been growing concern among developers working with Claude Code, the development platform of AI company Anthropic. Users report reaching their usage limits much faster than before, even when performing relatively light tasks.
According to Anthropic, there has been no tightening of limits, but customers are responding to the expiration of a temporary year-end bonus.
The Register reports that on forums, Reddit, and in the official Claude Developers Discord channel, developers indicate that their token consumption has suddenly increased significantly. As a result, accounts would reach their maximum within minutes or hours, whereas similar tasks previously posed little problem. Some users claim that their available token space appears to have decreased by tens of percent, based on analyses of their own log files.
Anthropic refutes these conclusions. According to the company, the observed changes are the result of the termination of a temporary doubling of usage limits that applied around the holidays. Between Christmas and New Year’s Eve, customers were given extra capacity to utilize unused computing power, as many business customers were less active during that period. Now that the bonus has expired, the limits have simply returned to normal levels.
Incomplete usage statistics
However, developers experience this differently. Several users indicate that they reach their limit remarkably quickly without intensive coding work, for example when reading documentation or reviewing specifications. They also point to missing or incomplete usage statistics in the management console, which makes it difficult to determine exactly where the high consumption is coming from.
The concerns are not limited to individual subscriptions. There are also reports from business accounts that token limits are being reached faster than expected. This fuels speculation about possible cost savings at Anthropic, especially now that the company is reportedly working towards an IPO. Anthropic denies that there is a link between financial plans and the current usage policy.
Some discussions point to a possible technical cause. According to some users, a recent version of Claude Code is less efficient with tokens. Some developers report that reverting to an older software version reduces the problem, although there is no consensus on this. Anthropic says it is investigating these reports but has not yet identified any structural error in token handling.
It is noteworthy that complaints about usage limits at Claude are not new. Since the fall, there have been extensive discussions about the balance between subscription price and available capacity. The recent spike in reports seems to coincide mainly with the end of the holidays, making it difficult to distinguish between perception and actual change.
For now, it remains to be seen whether Anthropic will provide additional transparency or technical adjustments. For developers who use Claude in their daily workflow, it has become particularly clear how sensitive AI services are to changes in terms of use, especially when these are not immediately visible or measurable.