Model Usage Tracking
Per-completion token tracking and basic aggregation is available to all users.
Per-model usage reporting and filtering by tags is available to enterprise users via the Cost Tracking API. Contact sales (opens in a new tab) for more information.
Cost Tracking API
GET
/api/metrics/modelStats
Parameters
Name | Required | Description |
---|---|---|
organization_id | Yes | The organization to retrieve usage stats for |
tags | No | Comma separated list of tags to filter by. Can be provided multiple times. Completions will only be counted if all supplied tags are matched (AND ). |
Response
Returns an array of datapoints that represent the input, output, and total token counts per model per day for the trailing 30 days.
Sample response
{
"data": [
{
"metric": "tokens_per_model_per_day",
"timestamp": "2024-04-25T00:00:00.000Z",
"total_tokens": 700,
"input_tokens": 366,
"output_tokens": 334,
"model": "gpt-4-1106-preview"
},
{
"metric": "tokens_per_model_per_day",
"timestamp": "2024-04-25T00:00:00.000Z",
"total_tokens": 400,
"input_tokens": 200,
"output_tokens": 200,
"model": "gpt-4-turbo"
},
{
"metric": "tokens_per_model_per_day",
"timestamp": "2024-04-24T00:00:00.000Z",
"total_tokens": 7325,
"input_tokens": 6951,
"output_tokens": 374,
"model": "gpt-4-1106-preview"
},
{
"metric": "tokens_per_model_per_day",
"timestamp": "2024-04-24T00:00:00.000Z",
"total_tokens": 4137,
"input_tokens": 3787,
"output_tokens": 350,
"model": "gpt-4-turbo"
},
...
]
}