Errors and Status Codes

When upstream capacity is busy, retry first

If you see a 502, 429, or model-capacity error, click retry first, or send continue in a client that supports continuation.

If repeated retries do not resolve the issue, review the API access route descriptions and system status, then contact support if needed.

When calling the API, you may see status-code errors. This page groups the most common cases and their next steps.

1. Upstream capacity and concurrency

These errors indicate upstream capacity or concurrency pressure rather than a local API key configuration error.

HTTP 502Our servers are currently overloaded. Please try again later.

Upstream (OpenAI) limit

Cause: The upstream service is temporarily handling more work than it can accept.

Next steps:

• Retry: Click retry in the client, or send continue when the client supports continuation.
• If it keeps failing, check the API access routes and system status before trying again.

HTTP 429Concurrency limit exceeded for account, please retry later

Upstream (OpenAI) limit

Cause: The concurrency capacity for the current model or pool is temporarily full.

Next steps: Wait briefly before retrying and avoid repeatedly sending the same request at once.

ErrorSelected model is at capacity. Please try a different model.

Upstream (OpenAI) limit

Cause: The selected model is temporarily at capacity.

Next steps:

• Retry later.
• If the task allows it, choose another model after checking its model name, capability, and price.

2. Network and stream interruption

This class of error means the connection ended before the response completed.

HTTP 504stream error: stream disconnected before completion

Network interruption

Cause: The client-to-gateway stream disconnected before completion.

Next steps:

• Switch access routes: Use CC Switch to switch API routes quickly. After switching, restart Codex or start a new session so the new route configuration takes effect.
• Configure a proxy only when needed: Do not add a proxy when using a direct route. Configure one according to your network environment only when you choose an international route.
• If partial output was already received, keep it and use the client’s continuation action instead of replaying the full request.

3. Local configuration

For a 401 error, check the API token and client configuration.

HTTP 401Invalid API Key provided / Unauthorized

Client request

Cause: The API token or endpoint configuration was rejected.

Next steps:

1. Confirm the API key is complete and has no extra spaces.
2. Confirm the Base URL points to the Qiuqiu Token route rather than the official OpenAI endpoint.
3. If the configuration is correct but repeated attempts still fail, recreate the token and configure the client again.

Use the documented recovery paths

Check the API access routes, system status, and usage logs before opening a support question. CC Switch can generate and switch client routes without manually retyping the Base URL and API token.