Skip to content

Model names in your client

Your code or tool names a model once; your organisation decides what serves it. The names to use are the ones approved in the Control room, and the gateway lists exactly the ones your key may call, so a tool's model picker shows nothing you'd be refused.

Which name to type

The names on the Models page of the Control room, for example gpt-4o-mini, gpt-4.1, claude-sonnet-4.5 or gemini-3.5-flash. They're whatever your organisation set up: an admin may give a model an alias, and may route a name to another model with a routing rule. The name works in any request format: claude-sonnet-4.5 in an OpenAI SDK, gpt-4o-mini in the Anthropic SDK.

Listing the models your key may call

Format Call Answered with
OpenAI GET /v1/models (client.models.list()) OpenAI's list shape
Gemini GET /v1beta/models (client.models.list()) Gemini's list shape

The gateway answers these itself, from the key: only the models that key may call, by the names to send. Tools that fill a model picker from /v1/models (most chat interfaces and coding assistants) show the same list.

Which model answered

Header Meaning
x-vidai-model The model that answered, after aliases, routing and fallback
x-vidai-requested-model The name you sent, when a different model answered
x-vidai-routing-rule The routing rule that applied, if any
x-vidai-fallback, x-vidai-fallback-from A fallback model answered

Each SDK page shows how to read them; where a tool can't, Request logs shows the model asked for and the model that answered, for every call.

Errors

Code Status Meaning
model_not_found 404 No model by that name: check the spelling against /v1/models
model_not_allowed 403 The model exists, but your key may not call it
denied_by_routing_rule 403 A routing rule refuses this model for you

Good to know

  • Some tools refuse names they don't know, or pick behaviour from the name (prefixes such as openai/). Their pages say how to give them your organisation's names.
  • More on aliases and routing: Model names and aliases.