The usage statistics page displays the resource consumption of AI services within a specified time range. You can view the total tokens, free tokens, paid tokens, and the distribution of calls by model and API key on this page to understand the actual usage of AI services and assist in cost analysis and resource optimization.
Click the organization account name in the top navigation bar to go to the project page. Click Usage Statistics in the left navigation bar to go to the usage statistics page. The usage statistics page mainly consists of a filter section, a tab navigation section, a token usage statistics section, a model statistics section, an API key statistics section, and a request statistics section.
Tab navigation section
The top of the page provides two tabs: Tokens and Requests, which allow you to switch between different statistical views:
Tokens: View data related to token usage, including the total number, free tokens, paid tokens, AI tool token usage and consumption trends, as well as model statistics and API key statistics.
Requests: View data related to request calls, including the number of requests, average response time, and request trends. You can also view the data by tool and memory.
Filter section
The top of the page usually provides time range and object filter capabilities to help you view AI service usage data within a specific period.
Common filter options include:
Time range: For example, the last 7 days or a custom start and end time.
Object: For example, view usage data by project.
Projects: Switch projects to view usage for different projects.
By using these filter options, you can analyze AI service usage trends across different time windows and business scopes.
Token usage statistics section
The token usage statistics section summarizes the overall resource usage under the current filter conditions. It typically includes the following metrics:
Total Tokens: The total number of tokens consumed within the selected time range.
Free Tokens: The number of tokens consumed within the free quota.
Paid Tokens: The number of tokens consumed beyond the free quota or according to billing rules.
AI Tool Token Usage: The number of tokens consumed by AI tools.
Model: Switch models to view token usage for different models.
This section also usually provides a Token Consumption Trend chart to display token usage changes over time, helping you identify peak periods, fluctuation cycles, and abnormal consumption.
Model statistics section
The model statistics section allows you to view token usage by model to analyze the call distribution and resource consumption characteristics of different models.
This section typically displays the following information:
Model: The name of the currently called model.
Provider: The provider or source of the model.
Input Tokens: The number of tokens consumed for request input content.
Cached Input Tokens: The number of input tokens that hit the cache.
Output Tokens: The number of tokens consumed for model-generated content.
API Key statistics section
The API key statistics section allows you to view call data by API key to analyze usage differences among different applications, services, or teams.
This section typically displays the following information:
API Key Name: Used to distinguish different callers.
Token Usage: The number of tokens consumed by the API key within the statistics period.
Requests: The number of requests made by the API key within the statistics period.
Request statistics section
Click Requests in the tab navigation area to enter the request statistics view. This view displays the request call details of the AI services, helping you analyze metrics such as request frequency and response performance.
The request statistics section provides two subtabs at the top: Tools and Memory, which are used to view request data from different dimensions:
Tools: View the call details of AI tool-related requests.
Memory: View the call details of requests related to AI memory. After selecting the Memory tab, you can filter by specific Memory Space in the filter section.
This section typically displays the following metrics:
Requests: The total number of requests initiated within the selected time range.
Average Response Time: The average response time for all requests, in milliseconds (ms), used to evaluate service performance.
This section also provides a Requests Trend chart, which displays the trend of requests over time to help you identify peak and off-peak hours, as well as any abnormal fluctuations.
Applicable scenarios
The usage statistics page is applicable in the following scenarios:
View overall resource consumption: Quickly understand the overall usage of AI services within a specific time range.
Analyze model usage distribution: Evaluate the call frequency and token consumption differences of different models.
Track business call sources: Identify the actual usage of different applications or teams by API key.
Monitor request performance: View request statistics such as the number of requests and average response time in the statistics view to assess service performance and stability.
Assist in cost optimization: Combine free token and paid token data with AI tool usage to optimize model selection and call strategies.
Recommendations
We recommend viewing token consumption trends on a fixed schedule to promptly detect abnormal growth or peak fluctuations.
We recommend analyzing both model statistics and API key statistics together to avoid only focusing on total usage without considering specific sources.
We recommend creating independent API keys for different business systems to facilitate detailed usage analysis.
If the output tokens of certain models are consistently high, we recommend further evaluating the prompt design, call frequency, or model selection.
We recommend regularly monitoring the average response time in the Requests view to promptly identify performance bottlenecks or service anomalies.
We recommend examining the request status in the Tools and Memory subdimensions within the Requests view to understand the call characteristics and potential issues of different functional modules.