APIMODELSAPIMODELS

Claude API error 529 "Overloaded": what it means

Updated 2026-11-19

TL;DR Short answer: 529 overloaded_error means Anthropic's API is temporarily overloaded by traffic from all users. Your request was fine and it is not your rate limit. Retry with exponential backoff, switch to another Claude model (capacity is tracked per model), or call through a provider that retries on a second route.

Model page: Claude Sonnet 5.5 API — pricing, specs and the playground

Anthropic's error reference describes 529 overloaded_error in one line: the API is temporarily overloaded, and it can happen when the API sees high traffic across all users. That is the key difference from a 429. A 429 is about your account; a 529 is about Anthropic's capacity at that moment, which is why it tends to arrive in waves around busy hours and new model launches.

In Claude Code it appears as "API Error: Repeated 529 Overloaded errors" after its own retries run out. The Claude Code docs are explicit that a 529 is not your usage limit and does not count against your quota. Over the raw API it can also arrive in the middle of a streamed answer as an error event named overloaded_error.

What a 529 looks like

WhereWhat you see
Raw API, no streamingHTTP 529 with {"type": "error", "error": {"type": "overloaded_error", "message": "Overloaded"}} and a request id
Raw API, streamingHTTP 200 first, then an SSE event named error whose data has error.type overloaded_error
Claude CodeAPI Error: Repeated 529 Overloaded errors, after Claude Code has already retried up to 10 times
Official SDKsAn overloaded error raised after the SDK's own retries (two by default)

How to fix a 529, in order

1. Wait and retry with exponential backoff. A 529 is temporary. The SDKs back off automatically; in your own code, double the delay each time and add a little randomness so many clients do not retry in lockstep.

2. Switch model. Capacity is tracked per model, so if Sonnet is overloaded, Opus or Haiku may not be. In Claude Code use /model.

3. Check status.claude.com if it lasts more than a few minutes.

4. For unattended jobs, let Claude Code keep trying. Setting CLAUDE_CODE_RETRY_WATCHDOG=1 makes it retry 429 and 529 capacity errors indefinitely, which suits CI runs. CLAUDE_CODE_OVERLOADED_RETRY_BASE_DELAY_MS (default 500) and CLAUDE_CODE_OVERLOADED_RETRY_MAX_DELAY_MS (default 32000) tune the backoff.

5. Handle it mid-stream. If overloaded_error arrives as a stream event, keep the partial answer and ask the model to continue.

What a 529 is not

It is not a rate limit, so lowering your request rate will not help much, and it is not caused by your prompt, your key or your plan. It also does not mean your money is being spent: in Claude Code it does not count against your quota, and on a pay-per-token API a request that fails is not a completed response.

Why 529s are rarer through APIMODELS

On APIMODELS, every current Claude model (Fable 5.2, Opus 5.5, Sonnet 5.5, Haiku 5.5) is served through at least two upstream routes. If the first route returns a 429, a 5xx or fails to connect before the response starts, the request is sent to the next route automatically, and you only see the error when every route has failed. Failed requests are not charged. One limit we cannot get around: once a response has started streaming to you, a failure in the middle of it cannot be moved to another route, so it reaches you as a stream error event.

So a 529 from one route before the answer starts is normally absorbed by the next route without your code noticing. You can still see a 529 when every route is overloaded at the same moment, or as a stream error if the overload hits after the answer has begun. Point the Anthropic SDK or Claude Code at https://api.apimodels.app and keep the same model names.

Início rápido

cURL

curl https://api.apimodels.app/v1/messages \
  -H "x-api-key: $APIMODELS_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{
    "model": "claude-sonnet-5-5",
    "max_tokens": 1024,
    "stream": true,
    "messages": [{"role": "user", "content": "Summarise this error log in three bullet points."}]
  }'

Python

import anthropic

# Same Anthropic SDK, pointed at APIMODELS. The SDK already retries 408/409/429/5xx
# with exponential backoff (twice by default); raise it for batch jobs.
client = anthropic.Anthropic(
    base_url="https://api.apimodels.app",
    api_key="sk_...your_apimodels_key...",
    max_retries=4,
    timeout=600,
)

try:
    with client.messages.stream(
        model="claude-sonnet-5-5",
        max_tokens=1024,
        messages=[{"role": "user", "content": "Summarise this error log in three bullet points."}],
    ) as stream:
        for text in stream.text_stream:
            print(text, end="", flush=True)
except anthropic.APIStatusError as e:
    # e.status_code: 429 / 500 / 529 ...; the body carries error.type and a request id
    print("failed:", e.status_code, e.response.headers.get("request-id"), e.message)

Perguntas frequentes

What does Claude API error 529 mean?

529 overloaded_error means Anthropic's API is temporarily overloaded by high traffic across all users. Your request is fine and it is not your rate limit. Retry with exponential backoff, or switch to another Claude model, because capacity is tracked per model.

Why is Claude overloaded?

Because demand across all users exceeds the capacity Anthropic has for that model at that moment. It clusters around busy hours and right after new model launches. It is a capacity problem on Anthropic's side, so changing your prompt or key will not fix it.

Does a 529 count against my Claude Code usage limit?

No. The Claude Code error reference states that a 529 is not your usage limit and does not count against your quota. Claude Code retries it up to 10 times before showing "Repeated 529 Overloaded errors"; after that, wait or switch model with /model.

How long does a Claude 529 last?

Usually minutes, not hours, but Anthropic does not publish a figure. If it lasts more than a few minutes, check status.claude.com. For unattended jobs, CLAUDE_CODE_RETRY_WATCHDOG=1 makes Claude Code keep retrying 529s instead of stopping.

Can I avoid 529s completely?

Not completely, because it reflects Anthropic's capacity. You can make it much rarer by retrying on a second route: on APIMODELS current Claude models run on at least two upstream routes, and a 529 before the answer starts is retried on the next route automatically. A 529 can still reach you if every route is overloaded at once.

Relacionado

Presente em
APIMODELS on Product HuntAPIMODELS featured on ToolifyAPIMODELS featured on There's An AI For That