> ## Documentation Index
> Fetch the complete documentation index at: https://docs.yourouter.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Rate Limits and Concurrency

> Handle model API throughput, provider limits, and retry behavior.

YouRouter is designed to support production model API traffic through one integration. For model calls, you should still build standard retry and concurrency handling because upstream model providers can enforce their own limits.

## What to Expect

| Area              | Behavior                                                                                                         |
| ----------------- | ---------------------------------------------------------------------------------------------------------------- |
| YouRouter gateway | Does not require a separate per-endpoint integration for each model provider.                                    |
| Provider limits   | The upstream provider may return rate limit or concurrency errors.                                               |
| Automatic routing | Omitting `vendor`, or using `vendor: auto`, lets YouRouter route to available providers for the requested model. |
| Pinned provider   | If you set `vendor`, your request depends on that provider's availability and limits.                            |

## Recommended Retry Pattern

For model API calls, retry transient failures with exponential backoff.

```text theme={null}
1st retry: wait 1 second
2nd retry: wait 2 seconds
3rd retry: wait 4 seconds
then stop or move to a fallback path
```

Use retries for:

* `429` rate limit or concurrency responses
* `500` gateway or provider errors
* temporary network failures

Do not retry immediately in a tight loop. That can make provider-side limits worse.

## Concurrency Tips

* Start with conservative concurrency in production and increase gradually.
* Use `vendor: auto` unless your integration requires a specific provider.
* Keep model IDs configurable so you can switch models without code changes.
* Log the request timestamp, model, vendor mode, and request ID for troubleshooting.
* For streaming responses, treat dropped connections as retryable only if your application can safely restart the request.

## Example: Automatic Routing

```bash theme={null}
curl https://api.yourouter.ai/v1/chat/completions \
  -H "Authorization: Bearer $YOUROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4o",
    "messages": [
      {
        "role": "user",
        "content": "Reply with exactly: connected"
      }
    ]
  }'
```

## Example: Provider Pinning

```bash theme={null}
curl https://api.yourouter.ai/v1/chat/completions \
  -H "Authorization: Bearer $YOUROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -H "vendor: openai" \
  -d '{
    "model": "gpt-4o",
    "messages": [
      {
        "role": "user",
        "content": "Reply with exactly: connected"
      }
    ]
  }'
```

For routing behavior, see the [Router guide](/guides/router). For request fields, see [Create Chat Completion](/api-reference/chat/create).