Because of a collection of developments — Anthropic declining to adjust to sure US authorities requests, and OpenAI stepping in to fill that position — two vital shifts occurred within the AI market nearly concurrently. ChatGPT took a wave of adverse public suggestions, and Claude obtained propelled to the highest of the charts.
A few days in the past, I made a decision to see what all the excitement was about. I tasked Claude, ChatGPT, and Gemini with constructing a simulation from the identical immediate. Claude’s outcomes have been so clearly superior that I canceled my ChatGPT subscription on the spot and signed up for Claude Professional as a substitute.
The joy was actual. With Claude Code and Claude Cowork each releasing across the identical time, it felt like the proper second to make the bounce. I moved my workflows over, dumped a bunch of paperwork and spreadsheets into it, prompted away, and obtained wonderful outcomes. Then I prompted once more — and obtained an error so heartbreaking it ruined my evening.
Claude has a a lot stricter quota
As a result of it meters compute, not messages
What greeted me, shortly after that second immediate, was a message telling me I had exceeded my quota. And in contrast to ChatGPT, whenever you exceed your quota on Claude, you exceed it for good. You are out. There is not any cheaper fallback mannequin to drop all the way down to. You are locked out of even the fundamental Haiku mannequin till the timer resets.
The saving grace is that limits reset each 5 hours quite than each month. Nevertheless it was nonetheless a intestine punch — particularly provided that I had simply paid for a Professional subscription.
Trying into it extra, the rationale turns into clear: Claude makes use of a wholly completely different methodology to calculate utilization in comparison with its rivals. ChatGPT and Gemini each have quota limits, however they’re structured in a means that finally ends up being extra forgiving to the typical person. OpenAI, for example, units limits based mostly on the variety of messages despatched. Plus customers get 300 messages per 3 hours with GPT-5.3, or 3,000 messages per week with GPT-5.4 Pondering. You can also make these messages as lengthy, brief, or document-heavy as you want. The three hundredth message is the one which will get you.
Anthropic makes use of compute-based metering as a substitute. It would not depend messages — it measures the precise computational work your requests generate. In case you oversimplify it, consider it like tokens: each doc hooked up, each lengthy response generated, each back-and-forth in a rising dialog thread provides to the tally.
The sensible consequence is stark. You would ship 300 informal messages to ChatGPT and by no means hit your wall. With Claude, two messages that contain a big spreadsheet and the Opus 4.6 mannequin could be sufficient to exhaust your quota totally. The flip facet can be true: in case your prompts are light-weight and centered, you would possibly squeeze out greater than 300 messages from Claude. However that is chilly consolation whenever you’re mid-workflow and immediately locked out.
You aren’t getting limitless utilization with any of the Claude plans
How did I burn via my quota so quick?
Picture by Amir Bohlooli. NAN.
I hadn’t learn the advantageous print. I assumed what most individuals assume: free tier is restricted, Professional is limitless. That is roughly how ChatGPT works in observe — not technically limitless, however the limits are set excessive sufficient that the majority customers by no means see them. I by no means had. Seems, Claude’s ceiling is way decrease and far more noticeable.
There’s additionally a significant distinction in how the 2 platforms deal with hitting that ceiling. ChatGPT degrades gracefully: exceed your quota, and it bumps you all the way down to a less expensive mannequin you may maintain utilizing till issues reset. Claude would not do this. In case you’re out, you are out. Shut the app, go contact grass, and wait.
Even Claude’s Max plans — which are available 5x and 20x tiers — aren’t limitless. They only multiply your baseline allowance by that issue. Claude can be notably opaque about what the precise limits are. They’re dynamic and shift based mostly on time of day, general system load, and utilization patterns. You will not discover a clear quantity on the pricing web page.
Here is what burned via mine so quick: I had Claude Cowork put in, and one in every of my ongoing objectives has been a clear integration between Obsidian and an LLM. I would tried pairing a neighborhood mannequin with Obsidian and hit bother. I would paired Obsidian with NotebookLM with first rate outcomes, however NotebookLM is not one thing you may run domestically. So I dumped chunks of my Obsidian vault into Claude and began asking questions. I obtained nice outcomes and saved going. And since I had simply subscribed, I used to be naturally working essentially the most highly effective — and most compute-hungry — mannequin out there: Opus 4.6. Anthropic describes it because the mannequin for formidable work. All my work is formidable work, clearly.
I wasn’t 4 messages into the dialog earlier than I hit the quota wall. That was it for Claude. Or so I believed.
Good habits make your quota final for much longer
And so they’re habits value having anyhow
Amir Bohlooli / MUO
When you perceive how Claude counts utilization, a number of changes make a major distinction — they usually’re the identical instincts you’d develop utilizing a neighborhood LLM.
The important thing psychological mannequin is context window price. Each time you ship a message in an ongoing dialog, Claude would not simply course of your new query — it re-reads the whole dialog historical past, together with all hooked up paperwork, earlier than producing a reply. Claude is processing that PDF and spreadsheet you hooked up in message one once more on message two, three, 4, and each message after. Mixed with a premium mannequin, this provides up startlingly quick.
A couple of habits that helped me use Claude higher:
Swap down from Opus. Since hitting that first wall, I’ve moved totally to Sonnet for day-to-day work. It is greater than succesful for many duties, and the compute price is dramatically decrease. Save Opus for the issues that truly want it.
Begin recent conversations typically. When you’re carried out with a activity, open a brand new chat quite than tacking on unrelated questions. Do not let context accumulate unnecessarily.
Clear your information earlier than attaching it. In case you’ve obtained a big uncooked spreadsheet, import it as soon as, have Claude produce a cleaned or condensed model, after which work from that. Attaching an enormous uncooked file to each follow-up query is among the quickest methods to eat via your quota.
Do not deal with Claude like a chat interface. The conversational UX makes it really feel like a messaging app, but it surely would not invoice like one. Consider it extra like knowledgeable service that fees by the hour — be intentional about what you convey into the room.
Claude additionally gives a quota extension when you hit your restrict. You’ll be able to flip it on and proceed on a pay-as-you-go foundation, however utilization is billed at the usual API charges, that are pretty costly.
Would I nonetheless suggest Claude Professional?
Sure. Claude remains to be essentially the most succesful mannequin I’ve used, particularly for extra technical work like coding. Pair that with Claude Cowork, and it is in a wholly completely different league from its competitors.
However, it does require a distinct relationship than ChatGPT. You’ll be able to’t simply fireplace and overlook. The quota limits are actual and the encircling transparency is poor. However, should you go in figuring out that and regulate accordingly, Claude Professional is value it. Simply do not spend your first evening throwing your whole Obsidian vault at Opus 4.6 and anticipating it to be advantageous.
