Manage Model Metadata
Register AI models with structured metadata in an AI Gateway instance to build a catalog that Model API and Route configurations can reference.
Prerequisites
-
You have created an AI Gateway instance. For more information, see Create an AI Gateway Instance.
-
At least one provider (preset or custom) is available in the instance. Models require a bound provider. For more information about managing providers, see Manage Providers.
Introductions
A model metadata entry describes a specific AI model under a provider. It includes:
-
Basic: Provider, name, source URL.
-
Credits: Price of Credit per Token of input/output/cache.
-
Performance: Maximum input/output/total Token, input/output models, capabilities.
-
Path: API path and protocols.
Requirements
-
If no provider is selected, submission is blocked.
-
If no name is entered, submission is blocked.
-
Credits consumption fields must be non-negative numbers with up to two decimal places.
-
Token limit fields must be non-negative integers.
-
Available paths must start with
/.
Manage Model Metadata
Log on to the AI Gateway console to manage model metadata. In the left-side navigation pane, select your AI Gateway instance and choose Model Metadata Management, then click the Model Metadata tab.
View the model list
On the Model Metadata tab, the model list displays the following columns:
|
Column |
Description |
|
Model Name |
Display name of the model |
|
Credits Info |
Credits per 1M tokens for input, output, and cache tokens |
|
Total Token Limit |
Maximum total tokens the model supports |
|
Capabilities and Performance |
Supported capabilities and their ratio to the total capability list |
|
Provider |
Bound provider name |
|
Updated At |
Last update time (format: YYYY-MM-DD HH:mm:ss) |
|
Actions |
View, Edit, Delete |
Create a Model Metadata
-
On the Model Metadata tab, click Create Model Metadata.
-
In the Create Model Metadata panel, configure the fields organized under collapsible sections:
Basic Information
Parameter
Type
Required
Default
Description
Provider
Select
Yes
—
Select an available provider from the current instance
Model Name
Input
Yes
—
Display name of the model
Credits Information
Parameter
Type
Required
Default
Description
Input Token Cost
Number (decimal)
No
0
Credits per 1M tokens, up to two decimal places
Output Token Cost
Number (decimal)
No
0
Credits per 1M tokens, up to two decimal places
Cache Token Cost
Number (decimal)
No
0
Credits per 1M tokens, up to two decimal places
Capabilities & Performance
Parameter
Type
Required
Default
Description
Max Input Tokens
Number (integer)
No
—
Maximum number of input tokens
Max Output Tokens
Number (integer)
No
—
Maximum number of output tokens
Max Total Tokens
Number (integer)
No
—
Total token limit
Input Modelity
Checkbox (multi-select)
No
—
Select from: text, image, audio, video
Output Modelity
Checkbox (multi-select)
No
—
Select from: text, image, audio, video
Capabilities (checkboxes):
-
System Message
-
Structured Output
-
Function Calling
-
Tool Choice
-
Prefix Completion
-
Prompt Cache
-
Computer Use
-
PDF Input
-
Vision
-
Image Embedding
-
Streaming Output
-
Parallel Function Calling
-
Web Search
-
Reasoning
-
URL Context
Available Paths
Parameter
Type
Required
Description
Path
Input
Yes
For example,
/v1/chat/completions. Must start with/Protocol Type
Select
Yes
Select from: OpenAI Compatible, Anthropic Compatible, Custom
-
-
Click OK to submit.
Edit a Model Metadata
-
On the Model Metadata tab, find the target model and click Edit in the Actions column.
-
Modify the fields as needed. The editable fields are the same as in the creation form.
-
Click OK to save changes.
Renaming a model does not automatically update existing Model API and Route configurations that already reference it, because the model name is stored as a string in those configurations. A confirmation prompt appears when you change the name.
Delete a Model Metadata
-
On the Model Metadata tab, find the target model and click Delete in the Actions column.
-
Confirm the deletion in the dialog.
-
After successful deletion, the model is removed from the list.
Deleting a model removes it from the candidate list but does not affect model name strings already saved in Model API and Route configurations.