I don’t settle Claude Pro vs Max on price. I settle it on whether waiting and other tools can cover what Pro falls short on.
I’ve hit the limit several times using Claude Code on Pro. I still haven’t upgraded to Max.
I’ve never used Max or usage credits, so the Max figures here come from the official docs.
On this page
What’s actually different between Claude Pro and Max for developers?
For developers, the biggest difference is usage. Max gives you 5x or 20x Pro’s usage per session, and both plans include Claude Code.
| Plan | Monthly price | Session usage vs Pro | Claude Code |
|---|---|---|---|
| Pro | $20 ($17/mo billed yearly) | Baseline | Included |
| Max 5x | $100 | 5x | Included |
| Max 20x | $200 | 20x | Included |
These values are from the official pricing page and the Max plan help article as of October 20, 2026.
The limit has two layers: a session that runs in 5-hour windows, and a weekly limit.
When do you hit Claude Pro limits in Claude Code?
I hit the limit on days when I used Claude Code intensively for several hours. When that happened, I usually waited 3–5 hours, and it cleared the same day.
Three kinds of work got me there:
- Refactoring an existing repository, reading through it and editing many files
- A loop of running tests, fixing failures, and running them again
- Days of long code reviews and design reviews back to back
On a normal workday afternoon, /usage shows the 5-hour session at 30–60%, with 1.5 to 3 hours left until reset. The weekly number is in the 40–70% range around midweek.
That’s my usual range, not today’s value. On any given day, /usage is the number to go by.
I don’t remember the weekly limit ever blocking me for days.
The most frustrating moment came one night. I was doing a refactor across several files and hit the limit right when I thought I was almost done.
Getting cut off with all that context built up felt worse than not being able to start at all.
It wasn’t urgent, so I wrote down the remaining work and waited.
Do Claude chat and Claude Code share the same limit?
Yes. I confirmed it in the official help article. So on a day of heavy web chat, Claude Code can have less room.
Claude Code is what hit the limit first for me. Web or app chat never hit it first.
In chat, I ask something short and it’s over. With Claude Code, one instruction turns into a long run of exploring, editing, testing, and fixing again from failure logs.
On one afternoon after a long web conversation with a big document pasted in, Claude Code felt like it hit the limit earlier than usual. That’s only a feeling, not confirmed.
Still, on days like that, I start by leaving room for coding.
Will an API key make Claude Code bill you instead of using the subscription?
It can, as far as I know, so I don’t keep my key exported in my shell profile. If ANTHROPIC_API_KEY is set in the environment, Claude Code can run on API billing instead of your subscription.
I put the key only in scripts that need the API. Everything else uses the subscription through the local login.
If the login gets tangled, I first run /status to see which account and method it’s connected with.
How should you read Max 5x and 20x?
5x and 20x are multipliers on Pro’s per-session usage. The Max help article doesn’t say how many times larger the weekly limit is.
So I don’t think you should read 20x as a guarantee of 20 times the weekly usage. That’s my interpretation of the doc’s wording.
It took me a while to tell the 5-hour limit from the weekly one, and that happened on other tools.
In Codex, I took the 5-hour window to mean five hours of actual use. But if the weekly limit runs out, you can’t use it even with time left in the 5-hour window. And with very heavy work, the 5-hour limit fills first even when the weekly one has room.
In Antigravity, I hit the weekly limit once. Usage didn’t come back even after the 5-hour refresh. Only after checking /usage did I see it was the weekly limit.
Claude has the same two-layer structure, so /usage is how I check. When I hit the limit, I first read whether the session or the weekly column filled up.
The 4 steps I go through before upgrading to Max
Before upgrading to Max, I check four things in order. It comes down to one question: is the time I lose to Claude limits consistently worth more than what Max would cost me extra each month?
- Can it wait? If it’s not urgent, I pick it up again 3–5 hours later.
- Can another tool take it? Urgent implementation goes to Codex.
- Can I cut overlapping subscriptions? I’ve already cut the overlapping AI subscriptions I was paying for.
- Does Claude still stop my work? That’s when I upgrade.
The choice isn’t a two-step move from Pro to Max. There are five options: waiting, another tool, usage credits, usage bundles, and Max.
According to the official help article, usage credits are billed at API rates after you use up your included limits, and you can set a monthly cap.
Handing work to another tool has a cost too. I had to re-explain to Codex the current goal, the scope of changes, the failing tests, and what not to touch. Codex also had to read through the repository and the diff from scratch.
So I split the work. Codex usually, Claude when uncertainty grows, and both for important changes.
What would make me upgrade to Max?
My line is 3 or more days a month when the limit pushed back my schedule, or more than 10 hours of total waiting in a month. I’m not there right now.
These numbers are a criterion I set, not a measurement.
I’m not holding off because Max is expensive. When I hit the limit, I can write down the remaining work and wait, or hand it to Codex. That’s enough to cover what Pro falls short on.
My evidence is my /usage log. I open it once a day, note the session and weekly numbers, and review them every 1–2 weeks.
That shows me the difference between heavy days and normal days, so I use it as the basis for the decision.
If I reach that line, I plan to start by trying Max 5x for a month. Based on the official prices, the extra cost is about $80, $100 minus $20.
I’ve moved up to a higher plan and come back down before, on other services: Gemini Ultra and ChatGPT Pro 500.
References
- Claude pricing: Pro and Max prices, Claude Code included, 5-hour and weekly limit structure
- What is the Max plan?: 5x and 20x prices and per-session multipliers
- Using Claude Code with your Pro or Max plan: shared limits with chat, API billing when an API key is set
- Manage usage credits for paid Claude plans: billed at API rates after included limits, monthly cap
Related posts
- Codex “Error Running Remote Compact Task: Ran Out of Room” Fix
- Codex CLI npm install Fails? It Was the Sandbox, Not npm
- Claude Outage Fallback: Request Failed, Files Still Changed
I first wrote this in Korean on my blog: https://jjeongil.tistory.com/3556.