Billing Units

Understand how text, image, audio, and video models are billed differently.

By modality

ParameterTypeRequiredDescription
Text / chattokensrequiredUsually billed by input, output, cached, and sometimes reasoning tokens.
Embeddingstokens or input unitsrequiredUsually based on input text tokens or model-specific input units.
Imageimages / resolution / qualityrequiredOften billed per generated image, resolution, quality, edit mode, or model unit.
Audioduration / characters / tokensrequiredMay depend on input audio duration, generated speech length, characters, or tokens.
Videoseconds / clips / qualityrequiredCommonly billed by generated duration, clip count, resolution, or quality tier.

How usage is reported

The console shows billable usage and charges together with request details such as model, status, and latency. Because reporting units differ by modality and model, review both the usage amount and the unit shown for each record.