Change a Model ID of an LLM Provider

Endpoint

POST /apiops/projects/{projectName}/llm-providers/{providerName}/rename-model/

Authentication

Requires a Personal API Access Token.

Header

Authorization: Bearer YOUR_TOKEN

Request

Headers

Header	Value	Required
Authorization	Bearer {token}	Yes
Content-Type	application/json	Yes

Path Parameters

Parameter	Type	Required	Description
projectName	string	Yes	Project name
providerName	string	Yes	LLM provider name

Request Body

{
  "from": "gpt-4o-mini",
  "to": "gpt-4o-mini-2026"
}

Field	Type	Required	Description
from	string	Yes	The current model ID on the provider
to	string	Yes	The new model ID. It must differ from from

Response

Success Response (200 OK)

The call also returns 200 when the old model is kept. Read completed and leftovers to see the outcome.

{
  "completed": true,
  "moved": [
    "API Proxy: chat-gateway",
    "Policy Group: shared-guards",
    "Limit Plan: engineering-daily"
  ],
  "redeployed": [
    "API Proxy: chat-gateway"
  ],
  "leftovers": [],
  "unreachedEnvironments": [],
  "warnings": {
    "refs": ["API Proxy: chat-gateway", "Policy Group: shared-guards"],
    "drafts": [],
    "conditionValues": [],
    "knowledgeBases": [],
    "limitAssignments": [],
    "variableModels": [],
    "limits": ["Limit Plan: engineering-daily"]
  }
}

A response where the old model is kept:

{
  "completed": false,
  "moved": [
    "API Proxy: chat-gateway"
  ],
  "redeployed": [
    "API Proxy: chat-gateway"
  ],
  "leftovers": [
    { "label": "API Proxy: batch-gateway", "reason": "DRAFT" },
    { "label": "Limit Assignment: team-a", "reason": "LIMIT_TARGET" }
  ],
  "unreachedEnvironments": ["test-env"],
  "warnings": {
    "refs": ["API Proxy: chat-gateway", "API Proxy: batch-gateway"],
    "drafts": ["API Proxy: batch-gateway"],
    "conditionValues": [],
    "knowledgeBases": [],
    "limitAssignments": ["Limit Assignment: team-a"],
    "variableModels": [],
    "limits": []
  }
}

Response Fields

Field	Type	Description
completed	boolean	true when the new model is listed, all references moved, and the old model is removed. false when the old model is kept; leftovers says why
moved	array of string	The records that now name the new ID, for example API Proxy: chat-gateway
redeployed	array of string	The API proxies and proxy groups that were running and were deployed again
leftovers	array of object	The records that stop the removal of the old model
leftovers[].label	string	The record, for example API Proxy: batch-gateway
leftovers[].reason	string	Why the record stops the removal. See the table below
unreachedEnvironments	array of string	Environments that the change could not reach and that do not run the model. They do not stop the change. They appear as failed in Deployment Results
warnings	object	What the change moves, as the Manager screen shows it before the change
warnings.refs	array of string	Where the model is named
warnings.drafts	array of string	API proxies and proxy groups with undeployed changes
warnings.conditionValues	array of string	Conditional routes that compare the model text with the old ID. They are not changed
warnings.knowledgeBases	array of string	Knowledge bases that name the model. The stored vectors are not computed again
warnings.limits	array of string	Limit plans and assignments. The running limit window may start again
warnings.limitAssignments	array of string	Limit assignments that target the model. They do not move. End them and create ones for the new model
warnings.variableModels	array of string	References that take their model ID from a variable (${...}, #{...}) that may name this model in an environment. They are not rewritten. The old model stays only while a value needs it

Leftover Reasons

Reason	Meaning	What to do
PROVIDER_DEPLOY_FAILED	The provider could not be sent to the gateways	Resolve the deploy error, then repeat the call
DRAFT	The proxy or group has undeployed changes, so it was not deployed. It runs the old ID	Deploy it, then repeat the call
DEPLOY_FAILED	The deploy failed. It runs the old ID	Deploy it again, then repeat the call
LIMIT_TARGET	A limit assignment targets this model. A target cannot change	End the assignment and create one for the new model
LIMIT_NOT_WRITTEN	The limit could not be written, for example a managed plan	Change the limit by hand
NOT_WRITTEN	The record could not be written	Repeat the call. Change the record by hand if it fails again
VARIABLE_MODEL	A reference takes its model ID from a variable. Apinizer resolves the variable in each environment and skips an environment with no value. In at least one environment the value reaches the old model and would reach no model once the old one is removed, or no environment knows a value (a request value #{...} counts as unknown). The old model stays and completed is false	Check the value of the variable in each environment. Then delete the old model from the provider's model list. A variable that names another model, or a base name such as w that reaches the new row w:new, does not stop the change
REFERENCE_REMAINS	The record still names the old ID. For example, an entry of another provider has no model of its own and takes its ID from a target that would move	Change the record by hand

Error Response (400 Bad Request)

{
  "status": "FAILURE",
  "resultMessage": "Model 'gpt-4o-mini' is not listed in this LLM provider"
}

A 400 also comes when from or to is empty, when the two IDs are equal, and when the provider is not found.

Two more refusals write nothing:

• Access lists differ. The to ID is already a model of the provider, and its access list differs from the access list of the from model. An access list that is off counts as empty.

{
  "status": "FAILURE",
  "resultMessage": "Model 'gpt-4o-mini-2026' already exists with a different access list. Make the access lists the same, or use a new model id"
}

• Missing permission. You lack the permission to edit, deploy or manage something that uses the model. The message names the assets.

{
  "status": "FAILURE",
  "resultMessage": "You do not have the permission to change these, so the model id was not changed: API Proxy: chat-gateway, Limit Plan: engineering-daily"
}

Error Response (401 Unauthorized)

{
  "status": "FAILURE",
  "resultMessage": "Token is not valid!"
}

cURL Example

curl -X POST \
  "https://demo.apinizer.com/apiops/projects/MyProject/llm-providers/openai-primary/rename-model/" \
  -H "Authorization: Bearer YOUR_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"from":"gpt-4o-mini","to":"gpt-4o-mini-2026"}'

Permissions

• User must have AI_DEVELOPMENT + MANAGE permission in the project
• User must have the permission to change every affected asset: API proxies, policy groups, global policies, proxy groups, knowledge bases and limit plans, and to deploy the proxies and groups that were running. If one is missing, nothing is written.

Notes and Warnings

• Order of steps: The call adds the new model and deploys the provider, moves the references, deploys again what was running, and removes the old model last.
• Repeat is safe: Repeat the same call after you resolve the leftovers. Each step looks at what is stored, and the call finishes the rest. A repeat after the old model was removed only deploys the provider again.
• Undeployed changes: A proxy or group with undeployed changes is not deployed. It runs the old ID until its owner deploys it. Repeat the call after that.
• Conditions are not changed: A condition on the model text of the client keeps its value. A client that sends the old ID still matches it and reaches the new ID.
• Knowledge bases: The call changes the model ID that a knowledge base names. It does not compute the stored vectors again.
• Name-based: The provider is addressed by providerName. The records in the response are named, never identified by internal ID.

Related Documentation

• Update LLM Provider - Update a provider and its model list
• Get LLM Provider - Read a provider
• Changing a Model ID - The same change in the Manager screen
• LLM Providers API - Resource overview and permissions