Rate limits

Calibrate runs your evaluations using AI providers such as OpenRouter and Deepgram. By default, Calibrate uses its own API keys for these providers and pays for the usage, so the rate limits are lower. If you add your own API keys (BYOK) for these providers, your provider accounts pay instead, and the rate limits are removed.

Rate limit	Default	BYOK
Number of tests in one agent evaluation run	20	No limit
Rows in one speech-to-text evaluation	20	No limit
Rows in one text-to-speech evaluation	20	No limit
Rows in one simulation	20	No limit
Number of evaluator scores in one run on a labelling task	20	No limit
Total production traces scored	100	No limit
Production traces stored	50,000	50,000

Request a rate limit increase

If you want to continue using Calibrate’s default API keys but want your rate limits increased, reach out to us using this form.

Get started

Agents

Evaluators

Human alignment

Simulations

Speech-to-text

Text-to-speech

Workspaces for Teams

Reference

Request a rate limit increase