Usage Metering
The Token Usage page provides visibility into token consumption and associated costs generated by inference activity. Usage metrics are continuously captured to help monitor consumption patterns, spending, and efficiency across models, deployments, and endpoints.
The page includes multiple views that present token usage data at different levels of detail, enabling both high-level monitoring and deeper analysis based on access scope and selected filters.
Overview¶
The Overview view presents a high-level summary of token usage and cost metrics across models, deployments, and endpoints. It provides a consolidated snapshot to quickly assess overall consumption and spending trends.
At the Partner level, the overview shows:
- Aggregate token usage across organizations and users
- Token consumption across model deployments and endpoints
- Cost distribution based on input and output tokens
- Top consuming models and deployments
At the Developer level, the overview shows:
- Token usage specific to the logged-in user
- Token consumption for models and deployments accessible to that user
The overview highlights:
- Total spend across organizations
- Token usage and active models
- Rate per 1K tokens and efficiency metrics
- Token distribution by input and output
- Cost distribution by token type
- Top organizations by usage
- Model distribution across deployments
Filters¶
Filters can be applied to refine token usage data displayed in the overview.
Date Range and Time
Usage data can be filtered using predefined or custom date ranges with time selection:
- Predefined ranges: Last 24h, 7d, 30d, 90d
- Custom date range with configurable start date, end date, start time, and end time
Additional Filters
Usage data can be further filtered by:
- Organization
- Project
- Model Deployment
- Endpoint
- Cost Type (input vs output tokens)
Token Usage¶
The Token Usage view provides detailed insights into token consumption over time. This view focuses on analyzing token usage volume and associated costs across model deployments, helping track trends and identify consumption patterns.
The page displays key token metrics, including:
- Total tokens consumed
- Input tokens
- Output tokens
Token usage trends are visualized through time-series charts that show:
- Token Usage Cost: Cost trends over time, broken down by model deployment
- Token Count: Token volume trends over time, broken down by model deployment
Hovering over a data point in the charts displays detailed usage information for a specific timestamp. For example, hovering over a peak in the Token Usage Cost chart shows the cost contribution from each model deployment at that time, allowing comparison of usage patterns across deployments.
This view helps analyze how token usage and costs evolve across different models and deployments during the selected time period.
Model Analytics¶
The Model Analytics view provides model-level insights into token usage and cost distribution. This view focuses on identifying cost patterns, usage concentration, and optimization opportunities across model deployments.
The page highlights key metrics, including:
- Total cost across models
- Cost concentration across top models
- Highest cost–contributing model
- Average cost per model
The Model Insights section surfaces observations and recommendations based on usage patterns, such as high cost concentration or significant cost variance across models.
Usage trends are visualized through a time-series chart that shows:
- Usage Trend: Cost trends over time, broken down by model deployment
Hovering over a data point in the chart displays the cost contribution of each model deployment at a specific timestamp, enabling comparison of usage patterns across models.
The Model List section provides a detailed breakdown of:
- Individual models
- Cost contribution and percentage share
- Distribution of usage across models
This view helps evaluate model-level efficiency, identify high-cost models, and make informed decisions to optimize usage and costs.
