Switching to Claude? These 4 habits help you avoid hitting usage caps  | amznusa.com

Just switched to Claude? Me too, but while it’s been a pleasure migrating to an AI that feels more grounded and less emoji-happy than its competitors, it’s also surprisingly easy to get cut off in the middle of a task.

Users on Reddit, especially those who just made the switch from ChatGPT or Gemini, have been complaining bitterly about how quickly they’ve been blowing through Claude usage limits, sometimes after only a handful of prompts on Opus 4.6, the most powerful Claude model.

Update [3/27/2026]: Anthropic has confirmed it’s “adjusting” peak five-hour session limits for free, Pro and Max subscribers while leaving overall weekly limits unchanged.

Even subscribers to Claude Max, the priciest ($100 or $200 a month) Claude subscription that offers up to 20 times the usage limits of the $20-a-month Claude Pro plan, have been feeling the pinch, with some Max users saying they’re “hitting walls in 30-40 minutes.”

So, is Claude that much tighter with usage limits than ChatGPT and Gemini? Frustratingly, the answer is “it depends,” and part of the issue is that Claude treats its chats and context windows in fundamentally different ways than ChatGPT and Gemini do. 

Another big reason you may be blowing through your Claude quotas so quickly is Claude’s new agentic AI tools, such as Claude Cowork and Claude’s new “computer use” functionality. Unlike a straightforward back-and-forth chat, the AI agents employed by Claude Cowork and particularly Claude Code often split up to perform multi-step tasks, and the more complex and magical those tasks are, the more tokens they’ll burn.

In any event, there are ways to make the most of your Claude usage quotas (which reset every five hours, including an overall weekly cap). I’ve scoured the Claude support site and Reddit forums for tips, and these are four of the best I’ve found.

Don’t let Claude chats get too long

One of the biggest differences between Claude and its top competitors is how they handle their context windows–that is, the total volume of information the model can consider at one time.

While ChatGPT uses a rolling context window to keep lengthy chats manageable (at the expense of “forgetting” context at the very beginning of a chat), Claude does its best to keep its context window as large as possible. Indeed, Anthropic recently rolled out support for an enormous one-million token context window, perfect for allowing Claude Opus and Sonnet to chew over an entire novel, academic paper, or software codebase in one gigantic gulp.

Of course, the problem with such a big context window is that each time you take another “turn” in the chat, everything prior gets resubmitted to Claude, and if the conversation gets too long, each back-and-forth will cost an increasingly large amount of tokens. That’s how you could wind up tearing through half of your Opus allotment in just a handful of turns.

So if you’re a ChatGPT user who’s used to letting chat threads grow to epic lengths, stop what you’re doing and start creating fresh chats whenever possible.

Craft Claude prompts with care

As we’ve already seen, tokens are money when it comes to Claude, and if you wander into a Claude chat with only a vague idea of what you want to accomplish, your subsequent back-and-forths will chew right through your usage allotment.

That’s why it’s key to plan your initial prompt carefully. Be detailed about what you want, and if you have questions for Claude about the task, ask them all up front in a single batch.

And if you do get into an extended exchange with Claude, request a detailed summary when you’re done and drop it into a new chat (see above) before starting the implementation phase.

Don’t neglect Claude’s Projects

Similar to ChatGPT and Gemini, Claude has Projects, or workspaces where you can attach supplementary documents that get added to the context window. 

Here’s the killer feature of Claude projects: Adding a document to a project only costs you tokens the first time you submit it. After that, Claude will consider the cached document during each turn of your chat without charging you for additional tokens. 

So, remember what I just said about getting Claude to summarize a chat for use in a fresh chat? You could also attach that summary to a Claude project and make even better use of your precious tokens.

Choose your Claude model wisely

It’s easy to get used to Claude Opus 4.6’s ability to power through complex problems that stymie other, less capable models.

But while it can be tempting to use Opus for everything, it’s an absolute fiend when it comes to tokens. For that reason, you’re better off keeping Opus stowed until you really need it. 

A good way to think of Opus is like an architect who draws up the plans for a new home. Once all the planning is completed, bring in Sonnet for the grunt work, while Haiku, the fastest Claude model, is great for simple polishing steps, like proofreading.

Opus can also be a great closer, coming in at the very end of a project for a final pass. And if you find Sonnet is hitting a wall, you can bring in Opus to mop things up.