Token Usage

Every response includes a usage object with prompt_tokens, completion_tokens, and total_tokens. Cost is calculated as:

cost = (prompt_tokens × input_price) + (completion_tokens × output_price)

Prices are per 1M tokens — see the for per-model rates. Your current balance and burn rate are visible in the dashboard.

For streaming responses, a final usage chunk may arrive in the SSE stream depending on the model provider. Image and video calls are priced per generation, not per token.