Multi-Region

LLM API can route each request to a specific region — useful for data residency, latency, or distributing load.

Supported for Azure OpenAI, AWS Bedrock, and Google Vertex. Other providers ignore the region header.

Choosing a region per request

Send the X-LLMAPI-Region header with any chat-style request (/v1/chat/completions, /v1/responses, /v1/messages):

curl -X POST https://api.llmapi.ai/v1/chat/completions \
  -H "Authorization: Bearer YOUR_LLM_API_KEY" \
  -H "Content-Type: application/json" \
  -H "X-LLMAPI-Region: eu" \
  -d '{
    "model": "azure/gpt-4o-mini",
    "messages": [{"role": "user", "content": "Bonjour !"}]
  }'

The value is case-insensitive and whitespace is trimmed. If the header is omitted or the alias is unknown, the request silently falls back to the provider's configured default region (typically eu) — it never fails with a 4xx because of this header.

Canonical region aliases

Only this closed set is accepted. Provider-native strings such as eu-west-1, europe-west4, or us-east-2 are not accepted and are treated as missing.

Alias	Region	Azure OpenAI	AWS Bedrock	Google Vertex
`eu`	European Union	configured resource	`eu-west-1`	`europe-west4`
`us`	United States	configured resource	`us-east-1`	`us-central1`
`ap`	Asia-Pacific	configured resource	`ap-southeast-1`	`asia-southeast1`
`global`	Global / cross-region	—	configured	`global`

Whether a given alias resolves to a real region depends on the deployment's server-side configuration. For AWS Bedrock, the matching cross-region inference profile prefix (us., eu., global.) is applied automatically.

INFO

google-ai-studio is unaffected by the region header — only the Vertex variant of Google models honours it.

Confirming which region served the request

For requests routed through a multi-region provider, the response echoes the header back so you can verify which region was requested:

X-LLMAPI-Region: eu

The echoed value is the client-supplied alias (lower-cased, trimmed). If the alias fell back to the default, the echo still shows the original alias the client sent. The region actually used is recorded per-request in your Activity logs.

Responses from providers that are not multi-region capable do not include this header.