Skip to main content
You aren’t limited to Opper’s hosted models. Register your own provider deployment, for example a private Azure OpenAI deployment, with its API key. It then appears in the catalog alongside built-in models and gets the same routing, governance, and tracing.

​
Register a custom model

The quickest way is in the dashboard. Open Settings → Models → Custom in platform.opper.ai and add a model:
FieldWhat to enter
NameA label, e.g. “My Custom Model”.
IdentifierHow you’ll call it, e.g. example/my-gpt4.
TypeThe provider type (Azure, OpenAI-compatible, and so on).
CredentialsThe provider’s API key, token, or service-account JSON.
API URLYour endpoint, e.g. https://my-deployment.openai.azure.com/.
Save it, then call the model by its identifier like any other.
Custom models are scoped to your project. The identifier prefix (e.g., example/) is the namespace.

​
Register from the API or CLI

If you’d rather automate it: 
curl -X POST https://api.opper.ai/v2/custom-models \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer ${OPPER_API_KEY}" \
  -d '{
    "name": "example/my-gpt4",
    "provider": "azure",
    "model_id": "gpt4-production",
    "api_key": "my-api-key-here",
    "params": {
      "api_base": "https://my-gpt4-deployment.openai.azure.com/",
      "api_version": "2024-06-01"
    }
  }'

​
Model aliases

An alias maps a stable name to an ordered list of models — a primary plus fallbacks. The gateway tries them in order, so the alias doubles as a backup chain: if the first model is down or rate-limited, the call continues to the next. The easiest way is the platform under Settings → Models. To automate it: 
curl -X POST https://api.opper.ai/v2/model-aliases \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer ${OPPER_API_KEY}" \
  -d '{
    "name": "production/main",
    "fallback_models": [
      "anthropic/claude-sonnet-4-6",
      "openai/gpt-5.5",
      "mistral/mistral-large"
    ]
  }'
Use production/main in your code, then reorder or repoint the chain anytime — no integration changes, no downtime. Custom models you’ve registered can go into the chain too.

​
What’s next

Models

The full catalog of built-in models.

AI Gateway

How routing, model selection, and BYOK fit together.