Claude Usage Limit Errors | How to Fix and Avoid Them by Error Type
For users who suddenly encounter errors like 529, Usage limit reached, 500, or 429 on Claude.ai, this article explains the causes and solutions based on official Anthropic information. From the differences between the 5-hour session limit, weekly limits, and API rate limits, to step-by-step retry procedures you can try immediately during overload situations — everything is organized so you can recover without confusion.
529 (Overloaded) is a temporary error caused by server overload and is not related to the user's input or plan status. It typically resolves within seconds to minutes, so in most cases simply reloading the browser or resending the same request is sufficient.
For API users, implementing exponential backoff for retries is the official recommendation. If errors persist, check status.anthropic.com for any widespread outages. Error 500 also requires no changes to your request; for 429, wait the number of seconds specified in the retry-after response header.
"Usage limit reached" means you have hit either the 5-hour session limit or the weekly limit. You can check your usage and reset time in Settings > Usage. If you need access restored sooner, upgrading to a higher plan is an option.
目次 (36)
- Claude's Usage Limit Structure — Two Axes: 5-Hour Session Limits and Weekly Limits
- 5-Hour Session Limit — Check Usage in Settings > Usage
- Weekly Limits — Separate from Session Limits, Wait Until the Following Week to Reset
- "90% of session limit" Warning — Advance Notice to Complete Tasks Before Hitting the Limit
- Free Plan Limits — Officially Noted as "Usage limits apply," Specific Numbers Not Disclosed
- "Usage limit reached" — Blocked When Either the 5-Hour or Weekly Limit Is Hit
- When It Appears — Reaching Either Session or Weekly Limits
- When It Resets — Session Resets After 5 Hours, Weekly Resets at the Start of the Next Week
- Error 529 (Overloaded) — 3 Things You Can Do Right Now
- What Is Error 529 — overloaded_error, a Temporary Issue from Server Overload, Not the User's Fault
- Fix 1/3: Wait a Few Seconds to Minutes and Retry — Overload Typically Resolves Quickly
- Fix 2/3: Automate API Retries with Exponential Backoff
- Fix 3/3: Check the Anthropic Status Page for Widespread Outages
- Error 500 Internal Server Error — An Anthropic-Side Issue, No Need to Change Your Input
- What Is Error 500 — api_error, an Unexpected Error Inside Anthropic's Systems
- How to Handle Error 500 — Wait and Retry; If It Persists, Contact Support with the request_id
- rate_limit_error (API) — HTTP 429 Means RPM / Token Limit Exceeded
- What Is Error 429 — Exceeding Per-Minute Limits on Requests / Input Tokens / Output Tokens
- Using the Retry-After Header — Retrying Before the Specified Time Will Always Fail
- Rate Limits by API Usage Tier — Tier 1 at RPM 50, Tier 4 at RPM 4,000
- "Claude Is Thinking Forever" — What to Do When There Is No Response for a Long Time
- What Is Happening — Complex Tasks, Unstable Network, or Server Delays Are the Main Causes
- 5-Step Fix — Wait a Few Minutes → Reload → New Chat → Check Network → Shorten Prompt
- Usage Limits by Plan — Free / Pro / Max / Team / Enterprise Allowances
- Best Practices for API Users {#api-} — Backoff, Caching, and Gradual Ramp-Up
- Implementing Exponential Backoff — Increase Intervals: 1–2 sec → 2–4 sec → 4–8 sec
- Using Prompt Caching — Cached Tokens Do Not Count Toward ITPM Rate Limits
- Ramp Up Traffic Gradually — Sudden Spikes Can Trigger Acceleration Limits and Cause 429
- What to Do When You Hit a Limit — Quick Reference by Situation
- Frequently Asked Questions (FAQ) — Answers to 5 Common Questions About Error Handling
- When does the limit reset?
- Do unused allowances roll over to the following week?
- Can I increase my usage limit for free?
- What is the difference between error 529 and error 429?
- Are Claude.ai usage and API usage counted separately?
- Sources (Primary Information)
Claude's Usage Limit Structure — Two Axes: 5-Hour Session Limits and Weekly Limits
Claude.ai's web and mobile apps have two types of usage limits set by Anthropic.
5-Hour Session Limit — Check Usage in Settings > Usage
Users on paid plans (Pro / Max / Team / Enterprise) can send messages within a 5-hour session limit. You can check your current usage via the progress bar in Settings > Usage.
The exact number of messages allowed is not specified on Anthropic's official pages. It varies by plan (Pro / Max / Team / Enterprise) and is adjusted through Anthropic announcements.
Source: Anthropic Support: Usage and rate limits (accessed: 2026-04-23)
Weekly Limits — Separate from Session Limits, Wait Until the Following Week to Reset
Separate from session limits, there are also weekly usage limits. Weekly usage can also be checked in Settings > Usage. If you reach the weekly limit, you must wait until it resets the following week.
Source: Anthropic Support: Usage and rate limits (accessed: 2026-04-23)
"90% of session limit" Warning — Advance Notice to Complete Tasks Before Hitting the Limit
When you approach 90% of your session usage, a warning message may appear. This is an advance notice that you are about to reach your limit. The recommended response is to complete your tasks or organize your conversation at this point.
Source: Anthropic Support: Usage and rate limits (accessed: 2026-04-23)
Free Plan Limits — Officially Noted as "Usage limits apply," Specific Numbers Not Disclosed
Usage limits also apply to the Free plan. The official pricing page states "Usage limits apply," but the specific number of messages is not listed.
Source: claude.com/pricing (accessed: 2026-04-23)
"Usage limit reached" — Blocked When Either the 5-Hour or Weekly Limit Is Hit
When It Appears — Reaching Either Session or Weekly Limits
Messages such as "You've hit your limit" or "Usage limit reached" appear when either of the following conditions is met:
- Usage within the 5-hour session has reached its limit
- Weekly usage has reached its limit
When It Resets — Session Resets After 5 Hours, Weekly Resets at the Start of the Next Week
Key points for this section:
- Session limit: Resets after 5 hours have passed
- Weekly limit: Resets at the start of the following week
The exact reset time can be checked in the Settings > Usage screen.
Source: Anthropic Support: Usage and rate limits (accessed: 2026-04-23)
Error 529 (Overloaded) — 3 Things You Can Do Right Now
What Is Error 529 — overloaded_error, a Temporary Issue from Server Overload, Not the User's Fault
HTTP status code 529 is defined as overloaded_error, indicating that the API is temporarily overloaded. It can occur for all users during high-traffic periods and is not caused by the user's request content or settings.
Source: Anthropic API Documentation: Errors (accessed: 2026-04-23)
Fix 1/3: Wait a Few Seconds to Minutes and Retry — Overload Typically Resolves Quickly
This is the simplest and most effective fix. Error 529 is caused by temporary server congestion and usually resolves within seconds to minutes. If you are using Claude.ai, simply reload the browser; if using the API, resend the same request.
Fix 2/3: Automate API Retries with Exponential Backoff
API users can automate retries by implementing exponential backoff. For specific code examples, see the "Best Practices for API Users" section later in this article.
Fix 3/3: Check the Anthropic Status Page for Widespread Outages
If error 529 keeps occurring frequently, there may be a widespread outage on Anthropic's end. Check status.anthropic.com to see if an official incident has been reported. If it is a widespread outage, individual fixes will not help — you will need to wait for Anthropic to restore service.
Error 500 Internal Server Error — An Anthropic-Side Issue, No Need to Change Your Input
What Is Error 500 — api_error, an Unexpected Error Inside Anthropic's Systems
HTTP status code 500 is defined as api_error, indicating that an unexpected error has occurred inside Anthropic's systems. It is not caused by the user's input or request content.
Source: Anthropic API Documentation: Errors (accessed: 2026-04-23)
How to Handle Error 500 — Wait and Retry; If It Persists, Contact Support with the request_id
Key points for this section:
- Wait and retry after some time
- If the problem persists, note the
request_idincluded in the response and contact Anthropic support
rate_limit_error (API) — HTTP 429 Means RPM / Token Limit Exceeded
What Is Error 429 — Exceeding Per-Minute Limits on Requests / Input Tokens / Output Tokens
HTTP status code 429 is defined as rate_limit_error, indicating that the per-minute limit on API requests (RPM) or token counts (ITPM / OTPM) has been exceeded.
Source: Anthropic API Documentation: Errors (accessed: 2026-04-23)
Using the Retry-After Header — Retrying Before the Specified Time Will Always Fail
The response for a 429 error includes a retry-after header, which specifies how many seconds to wait before the next request. Retrying before this time will always fail.
Source: Anthropic API Documentation: Rate limits (accessed: 2026-04-23)
Rate Limits by API Usage Tier — Tier 1 at RPM 50, Tier 4 at RPM 4,000
The following are the standard rate limits (RPM = requests per minute) for the Messages API as listed in Anthropic's official documentation.
| Model | Tier 1 RPM | Tier 2 RPM | Tier 3 RPM | Tier 4 RPM |
|---|---|---|---|---|
| Claude Sonnet 4 series | 50 | 1,000 | 2,000 | 4,000 |
| Claude Haiku 4.5 | 50 | 1,000 | 2,000 | 4,000 |
| Claude Opus 4 series | 50 | 1,000 | 2,000 | 4,000 |
- For exact limits per model, please refer to the source. Detailed tables for ITPM (input tokens per minute) and OTPM (output tokens per minute) are also available on the same page (accessed: 2026-04-23)
"Claude Is Thinking Forever" — What to Do When There Is No Response for a Long Time
What Is Happening — Complex Tasks, Unstable Network, or Server Delays Are the Main Causes
After sending a message to Claude, sometimes the response does not appear and the "thinking" indicator continues for a long time. Possible causes include:
- Extended processing time due to complex tasks or long contexts
- Unstable network connection
- Server-side delays
5-Step Fix — Wait a Few Minutes → Reload → New Chat → Check Network → Shorten Prompt
Key points for this section:
- First, wait a few minutes — Tasks involving Extended Thinking may take longer to process
- Reload the browser — Refresh the page to check the conversation state
- Start a new chat — Effective when the issue is limited to a specific conversation
- Check your network connection — If the connection is unstable, try switching Wi-Fi networks or reconnecting
- Shorten the prompt and resend — Reorganize and resend conversations with a large amount of context
Usage Limits by Plan — Free / Pro / Max / Team / Enterprise Allowances
The following information is based on the Anthropic official pricing page as of 2026-04-23.
| Plan | Price (excl. tax, USD) | Usage Overview | Notes |
|---|---|---|---|
| Free | $0 | Limited (details not disclosed) | No credit card required |
| Pro | $20/month (monthly billing) $17/month equivalent (annual billing) |
More than Free | Claude Code access included |
| Max | From $100/month | 5x / 20x Pro | For heavy users |
| Team | $20/seat/month (annual billing, standard seats) | Varies by plan | Minimum 5 seats |
| Enterprise | Contact for pricing | Custom | May include API pay-as-you-go |
The exact number of messages is not stated on Anthropic's official pages, so it is described as "varies by plan and adjusted through Anthropic announcements." Max's "5x / 20x" is based on the official description relative to Pro ("5x the usage of Pro" / "20x the usage of Pro"). Source: claude.com/pricing (accessed: 2026-04-23)
Best Practices for API Users {#api-} — Backoff, Caching, and Gradual Ramp-Up
Implementing Exponential Backoff — Increase Intervals: 1–2 sec → 2–4 sec → 4–8 sec
When you receive a 429 or 529 error, retrying at fixed intervals will only make the situation worse. Use exponential backoff to progressively increase the time between retries.
- First failure: wait 1–2 seconds
- Second failure: wait 2–4 seconds
- Third failure: wait 4–8 seconds
- Set a maximum retry count and propagate the error upward if exceeded
The official Anthropic SDKs (Python / TypeScript) have automatic retry functionality built in.
Source: Anthropic API Documentation: Rate limits (accessed: 2026-04-23)
Using Prompt Caching — Cached Tokens Do Not Count Toward ITPM Rate Limits
With Claude's API, cached input tokens do not count toward the ITPM (input tokens per minute) rate limit.
- Actively cache system prompts, large context documents, and tool definitions
- Higher cache hit rates lead to dramatically improved effective throughput
- You can check your cache hit rate on the Usage page in Claude Console
Example: With an ITPM limit of 2,000,000 and an 80% cache hit rate, you can effectively process 10,000,000 tokens per minute.
Source: Anthropic API Documentation: Rate limits (accessed: 2026-04-23)
Ramp Up Traffic Gradually — Sudden Spikes Can Trigger Acceleration Limits and Cause 429
Sudden spikes in traffic can trigger "acceleration limits" and cause 429 errors. When going live in production, it is recommended to increase request volumes gradually and maintain a stable pattern.
Source: Anthropic API Documentation: Rate limits (accessed: 2026-04-23)
What to Do When You Hit a Limit — Quick Reference by Situation
Key points for this section:
| Situation | Recommended Action |
|---|---|
| Reached session limit | Wait for the 5-hour reset |
| Reached weekly limit | Wait for the reset at the start of next week |
| Frequently hitting limits | Consider upgrading to a higher plan |
| Hit limit on Free plan | Upgrade to Pro plan ($20/month) |
| Pro plan isn't enough | Consider Max plan (from $100/month) |
| Persistent 429 errors on API | Consider Prompt Caching / batch processing / tier upgrade |
Prices exclude tax, in USD. Source: claude.com/pricing (accessed: 2026-04-23)
Frequently Asked Questions (FAQ) — Answers to 5 Common Questions About Error Handling
Here are answers to the questions most frequently asked by readers.
When does the limit reset?
The session limit resets after 5 hours have passed; the weekly limit resets at the start of the following week. The exact time can be checked in Settings > Usage.
Source: Anthropic Support: Usage and rate limits (accessed: 2026-04-23)
Do unused allowances roll over to the following week?
As of 2026-04-23, no information about rollovers could be found on the official pages. For the latest information, please check the Anthropic Support Center. [Unconfirmed]
Can I increase my usage limit for free?
As of 2026-04-23, there is no information in official sources about a way to increase the Free plan's usage limit without paying. Upgrading to a paid plan is required to increase the limit.
Source: claude.com/pricing (accessed: 2026-04-23)
What is the difference between error 529 and error 429?
Key points for this section:
- 529 (overloaded_error): The entire server is overloaded. May affect all users.
- 429 (rate_limit_error): Your account's rate limit has been exceeded. Resolved by waiting and retrying.
Source: Anthropic API Documentation: Errors (accessed: 2026-04-23)
Are Claude.ai usage and API usage counted separately?
Yes. Usage under a Claude.ai subscription (Pro / Max / Team) and Anthropic API usage are separate systems. Using the API requires separate registration through Claude Console and setting up pay-as-you-go billing. For details, see Getting Started with the Claude API.
Source: claude.com/pricing (accessed: 2026-04-23)
Sources (Primary Information)
- Source — Plan pricing and usage limit overview (accessed: 2026-04-23)
- Anthropic Support: Usage and rate limits — 5-hour session limits, weekly limits, and Settings > Usage explanation (accessed: 2026-04-23)
- Anthropic API Documentation: Rate limits — Rate limit figures by tier, Prompt Caching, response headers (accessed: 2026-04-23)
- Anthropic API Documentation: Errors — HTTP error code definitions (400 / 429 / 500 / 529, etc.) (accessed: 2026-04-23)
