How can I check my API usage and limits with OpenAI?

To check your API usage and limits with OpenAI, you can use the OpenAI platform dashboard and programmatic methods. Start by logging into your OpenAI account at platform.openai.com. Navigate to the “Usage” section, which displays key metrics like API call volume, costs, and current rate limits. The dashboard provides a clear breakdown of usage by date, model type (e.g., GPT-4, GPT-3.5), and specific API endpoints. For example, you can see how many tokens were processed in the last 24 hours or track monthly spending against your account’s quota. This interface also shows your account’s rate limits, such as requests per minute (RPM) and tokens per minute (TPM), which vary based on your subscription tier and usage history.

For automated monitoring, use OpenAI’s API to fetch usage data programmatically. Send a GET request to the https://api.openai.com/v1/usage endpoint with your API key in the header. You can filter results by specifying a start_date and end_date in the query parameters. For example, a Python script using the requests library might retrieve data for the current month. Additionally, check rate limits directly from API responses: every call returns HTTP headers like x-ratelimit-limit-requests (max allowed requests per minute) and x-ratelimit-remaining-requests (remaining quota). These headers help you adjust your application’s request pacing in real time. If you hit a limit, the API returns a 429 status code, signaling the need to implement retries with exponential backoff.

Understanding your specific rate limits is critical. Free trial accounts typically start with lower limits (e.g., 20 RPM/150,000 TPM for GPT-4), while paid tiers scale these thresholds. You can request a limit increase via the OpenAI support panel in the dashboard, but approval depends on your usage history and adherence to policies. For cost tracking, the dashboard’s “Billing” section itemizes expenses per model, helping you optimize which models to use. For example, switching from GPT-4 to GPT-3.5-turbo for non-critical tasks reduces costs. Always test your application under expected loads to ensure it handles rate limits gracefully, using tools like logging or monitoring services to track API errors and usage trends over time.