Anthropic, the AI research company behind the popular Claude models, has announced new weekly rate limits for its AI coding tool, Claude Code. This move, set to go into effect on August 28, 2025, aims to rein in usage by a small segment of "power users" who are reportedly running the AI continuously, 24/7, in the background. It's a significant development, especially for developers who rely heavily on Claude Code for their daily workflows.

The company stated that these new limits are also designed to combat violations of Claude’s usage policy, specifically targeting instances of account sharing and the unauthorized reselling of access to Claude Code. This isn't just about managing computational resources; it's also about maintaining the integrity of their service.

These new weekly limits will apply to subscribers across Anthropic’s paid tiers: the $20-per-month Pro plan, and the more expensive $100- and $200-per-month Max plans. The announcement came via an email to subscribers and a post on X (formerly Twitter) on Monday, July 28.

It's important to note that Anthropic's existing usage limits, which reset every five hours, will remain in place. The new additions introduce two distinct weekly caps that reset every seven days. One is a general overall usage limit, while the other is specifically tailored for Claude Opus 4, Anthropic's most advanced AI model. For those on the Max plans, there's a silver lining: the option to purchase additional usage beyond the rate limit at standard API rates. So, if you're really pushing the boundaries, you still have options, albeit at an extra cost.

Anthropic has provided some estimated usage hours for its models under these new limits. For most Pro users, they can anticipate between 40 to 80 hours of Sonnet 4 through Claude Code per week. Max subscribers paying $100 a month will see 140 to 280 hours of Sonnet 4 and 15 to 35 hours of Opus 4. The top-tier $200-per-month Max plan offers 240 to 480 hours of Sonnet 4 and 24 to 40 hours of Opus 4.

Now, here’s where things get a bit murky. Anthropic previously touted its $200 Max plan as offering "20x more usage than the Pro plan." But if you crunch the numbers based on these new Claude Code hour estimates, it looks more like about 6x the usage compared to Pro. The company hasn't immediately clarified this discrepancy, though it's possible the 20x figure still holds true when measured by tokens or raw compute rather than just "hours." It's a detail many users will want more clarity on, I'd bet.

This isn't Anthropic's first rodeo with usage adjustments. Just weeks prior, the company quietly implemented rate limits for Claude Code without much fanfare, leaving some users scratching their heads. At the time, Anthropic acknowledged "issues" but kept mum on the specifics. Now we know why.

Claude Code has been a massive success, no doubt. "Claude Code has experienced unprecedented demand since launch," Anthropic spokesperson Amie Rotherham confirmed in an email to TechCrunch. But that success has come with its own set of challenges. The company's status page tells a story of frequent disruptions, with Claude Code experiencing partial or major outages at least seven times in the last month. It's not hard to connect the dots: continuous, 24/7 usage by a small percentage of power users likely contributes significantly to these service interruptions.

Rotherham also emphasized that the new limits are expected to impact "less than 5% of subscribers" based on current usage patterns, suggesting this is a targeted intervention rather than a broad crackdown. It makes sense, doesn't it? You don't want a few bad apples spoiling the experience for everyone.

Anthropic isn't alone in grappling with the immense computational demands of serving advanced AI models. It's a common refrain among AI model providers these days: everyone's racing to bring new AI data centers online just to keep up. Computational resources are simply very constrained across the board.

This struggle isn't unique to Anthropic's Claude Code. We've seen other AI coding tool providers revisit their pricing and usage strategies recently. Take Anysphere, the company behind Cursor. Back in June, they altered their $20-per-month Pro plan to curb power user abuse. Though, to be fair, they later apologized for poorly communicating those changes, which led to some user frustration and unexpected costs. Replit, another prominent AI coding tool, also made similar "effort-based pricing" adjustments in June. It seems the entire industry is figuring out how to balance user demand with sustainable resource allocation. It’s a tough tightrope walk, for sure.

For the vast majority of Claude users, Anthropic maintains that they “won’t notice a difference.” The goal, as stated by the company, is to “maintain reliable service broadly in the short term.” While these new limits might be a bummer for the small percentage of heavy users, they're framed as a necessary step to ensure a stable and consistent experience for the wider subscriber base.

Anthropic has also hinted at future solutions, stating in an email to Claude subscribers that it's committed to "supporting long-running use cases through other options in the future." What those options might be remains to be seen, but it suggests they're not abandoning the idea of continuous, high-volume AI usage entirely. Perhaps a new, even higher-tier plan, or dedicated enterprise solutions? Only time will tell, but it's clear the AI industry is still very much in a phase of rapid evolution, and pricing models are evolving right along with it.