model_create
ManagementCreate Model
Create a new model.
Args:
- data: required — model payload. Common fields: name, shortDescription, tags, thumbnail, privacy, type. Less-common training/parameter fields (concepts, epoch, parameters, negativePromptEmbedding, etc.) pass through via passthrough().
- team_id, project_id: required for OAuth callers
Returns: the created model record.
Example: "Create a new model" -> data={"name": "My Model", "type": "sd-xl-lora"}
open-world
Parameters
| Name | Type | Required | Description |
|---|---|---|---|
| data | unknown | ✓ | Model data for create. Common fields above; less-common training/parameter fields (concepts, epoch, parameters, negativePromptEmbedding, etc.) pass through unchanged via `.passthrough()`. See ModelCreateParams in the SDK for the full surface. |
| team_id | string | — | Team ID. Required if user belongs to multiple teams. |
| project_id | string | — | Project ID to scope the operation to. |
| response_format | enum(json | markdown) | json | Output format: 'json' for structured data, 'markdown' for human-readable text. |
Example Request
JSON
{
"data": {
"name": "Dragon Style v2",
"type": "flux.2-dev-lora"
},
"team_id": "team_abc123",
"project_id": "proj_xyz789"
}Example Response
JSON
{
"model": {
"id": "model_custom_new",
"name": "Dragon Style v2",
"status": "draft"
}
}Common Use Cases
- Create a new model placeholder before uploading training images
- Use the type from recommend_training to pick the right architecture