There are plenty of articles explaining how Claude Pro limits work. What’s missing is the actual cost of what to do the moment you hit one. The top six search results all carefully explain the “5-hour window and weekly cap,” then reach the same conclusion: “wait or upgrade.” This page adds a third option and compares all three on the same basis — cost.
How it works: limits count tokens, not messages
Limits use rolling windows, not a fixed monthly quota. The window starts with your first message, and that usage becomes available again after a set period. There’s also a separate weekly cap, which can be reached before the 5-hour window if you use a lot in a short time.
The key point is that usage is measured in tokens, not messages. With every exchange, the entire conversation history is sent again, so the same “one message” counts more when the conversation is longer. Having Claude read a file and make changes can use tens of thousands of tokens in a single step. That’s almost always why you “hit the limit so quickly”—it’s not a settings issue.
A common misconception: upgrading doesn’t reopen a window that’s already closed
Upgrading to a higher plan gives you more capacity in subsequent windows. It doesn’t reopen a window that’s currently closed. If you hit a limit in the middle of deadline-driven work, upgrading won’t solve the problem. Get this wrong, and you’ll pay more while your work remains stalled for the day.
Reset times aren’t fixed either. The window starts when you send your first message, so “what time it resets” varies from person to person. The resume time shown on the limit screen is the only accurate information.
Compare the three options by cost
| Option | Does it work immediately? | Best for |
|---|---|---|
| Wait for the window to reopen | No (you have to wait) | No deadline; you can wait a few hours |
| Switch to a higher-tier plan | No (only from the next window) | You hit the limit every day. You consistently need more capacity. |
| Switch to pay-as-you-go for that task only | Yes (immediately) | You need to finish today and hit the limit only a few times a month |
The third option doesn’t require canceling your subscription. While the next two lines are set, you’ll pay per token; remove the variables to switch back to your subscription.
# 制限に当たった作業だけを従量課金に逃がす 2 行。
# サブスクは解約しない — 変数がある間だけキー課金、消せば元に戻る。
export ANTHROPIC_BASE_URL=https://api.kunavo.com
export ANTHROPIC_AUTH_TOKEN=sk-kn-...
# Claude Code の既定モデルと opus / sonnet エイリアスは Anthropic の最新モデルを指す。
# sonnet は Kunavo が提供していない Sonnet 5.5 を呼んで 404 になるので、モデルを固定する。
# opus の固定先 Opus 5.5 は Claude Code v2.1.280 以降が必要（古ければ claude update）。
export ANTHROPIC_MODEL=claude-sonnet-5
export ANTHROPIC_DEFAULT_OPUS_MODEL=claude-opus-5-5
export ANTHROPIC_DEFAULT_SONNET_MODEL=claude-sonnet-5
# 補助的な内部呼び出しを一番安いモデルに寄せる
export ANTHROPIC_DEFAULT_HAIKU_MODEL=claude-haiku-4-5How much pay-as-you-go actually costs
Rates are loaded directly from the catalog (not entered manually). Per 1M tokens:
| Model | Input / output |
|---|---|
| Claude Haiku 4.5 | $0.70 / $3.50 |
| Claude Sonnet 5 | $1.40 / $7.00 |
| Claude Opus 5.5 | $2.80 / $14.00 |
Since unit prices are hard to put into perspective, let's convert them into the cost of one task. One step with 25,000 input / 1,200 output tokens costs about $0.0434 with Claude Sonnet 5, meaning $1 covers about 23 steps. If a task takes 20 steps to complete, it costs about $0.868 per task. Costs are deducted from your prepaid balance, so your bill is 0 for any month in which you do not use the service.
One calculation shows which option is cheaper: monthly price ÷ cost per task = break-even number of tasks. The detailed break-even calculation is in Claude Code pricing.
Three ways to hit limits less often in the first place
- Start a new conversation. Simply starting a new conversation between unrelated tasks can significantly reduce usage. One endless thread is the most expensive option.
- Choose the model based on the size of the problem. Do not use the largest model as the default — as shown in the table above, the unit prices for Claude Haiku 4.5 and Claude Opus 5.5 differ by approximately 4 times.
- Use caching. If you send the same context every time, caching can greatly reduce input usage (an explanation of prompt caching).
For Claude Code workflows (CLAUDE.md), permissions, and custom commands, see How to use Claude Code. For the detailed English version, see Claude Pro and Max limits.
Frequently asked questions
How do Claude Pro limits work?
Usage is managed through rolling windows, not a fixed monthly quota. The window starts when you send your first message, and capacity is restored after a set period. There is also a separate weekly cap, so heavy use over a short period can hit the weekly cap before the 5-hour window limit. Consumption is measured in tokens, not messages, so long conversations and large attachments count more even if they are each just one “message.”
If I upgrade to a higher plan, can I use it right away?
No. This is the point people most often misunderstand. Changing plans increases the limit from the next window onward; it does not open the window that is currently closed. If you hit the limit in the middle of work with a deadline, upgrading will not solve the problem. The only ways to continue right then are to wait or switch that task to a pay-as-you-go API key.
Why do I hit the limit so quickly?
In many cases, it is because the conversation is long. Each exchange resends the entire history so far, so the longer the conversation gets, the more each message consumes. Agentic use—having it read and edit files—is even heavier and can use tens of thousands of tokens in one step. Starting a new conversation with only the key points carried over works more reliably than changing any setting.
What can I do to keep going when I hit a limit without waiting?
Yes. You can switch to an API key just for that task without canceling your subscription. Token-based usage billing applies only while ANTHROPIC_BASE_URL and ANTHROPIC_AUTH_TOKEN are set; removing these variables switches you back to your subscription. As a guide, one step with 25,000 input / 1,200 output tokens costs about $0.0434 with Claude Sonnet 5 — $1 covers about 23 steps.
Can I check when my window resets?
The window starts when you send your first message; it doesn’t start at the same time for everyone. That means reset times vary, so there’s no fixed answer to “what time will it reset?” The limit screen shows when you can resume, and that’s the only accurate information.
Will switching to pay-as-you-go cost more?
It depends on how you use it. The calculation is simple: monthly price ÷ cost per task = break-even number of tasks. If a task takes about 20 steps, it costs about $0.868 per task. In months when you run more than that number of tasks, the subscription is cheaper; when you run fewer, pay-as-you-go is cheaper. You can also use both: keep your subscription for everyday work and switch to a key only when you hit a limit.