Claude Code gets throttled as Anthropic rolls out fresh usage caps

Claude Code, the AI-in-terminal utility developed by Anthropic and launched back in February, is getting updated usage limits following weeks of user complaints about being abruptly cut off. Many developers on the "$200/month Max plan" found their access blocked after just a few requests, with no explanation from the company.

In a recent thread posted to X, the AI lab explained that it has seen "unprecedented demand since launch," pointing to some of its heaviest users who were running the tool continuously in the background 24/7, with one person reportedly consuming tens of thousands of dollars in model usage on a single $200 subscription.

Anthropic also claimed that some users were violating its usage policy by sharing and reselling accounts, which impacts system capacity for everyone. These factors all led the company to announce new weekly limits that will be added on top of the existing five-hour caps, effective August 28. Max plan subscribers will have the option to buy additional usage at standard API rates if they hit their cap.

Here"s what the new weekly limits look like:

• Pro Plan ($20/month): An estimated 40 to 80 hours of usage with the Sonnet 4 model.
• Max Plan ($100/month): An estimated 140 to 280 hours with Sonnet 4 and 15 to 35 hours with the top-tier Opus 4 model.
• Max Plan ($200/month): An estimated 240 to 480 hours with Sonnet 4 and 24 to 40 hours with Opus 4.

Per TechCrunch, the company provided these hour-based estimates, noting that the actual numbers may vary based on the size of a project"s codebase. What"s interesting is how this new structure compares to the old marketing. Anthropic previously advertised its $200 Max plan as offering 20 times more usage than the Pro plan.

Based on these new hourly estimates, that multiple is now closer to six. It is possible the 20x figure still applies when measured in tokens or raw compute, but, according to TechCrunch, the company has not clarified that point.