← Pricing
PricingModels

Omni

Models that can return either text or video in one interaction.

How billing works

Omni models can return text or video. The API reserves for the highest known output case before the call so the account cannot be undercharged.

  • Text and video output use different rates.
  • The unused part of a reservation is returned after the provider reports actual usage.
  • The output mode is decided by the model, not selected as a separate checkout option.

Price in practice

The reservation protects the account before the model decides whether its response contains text or video. Settlement uses the measured outcome rather than leaving the maximum reservation in place.

For cost-sensitive flows, treat video-capable responses as potentially more expensive than text-only responses.

Rate card

ServicePriceHow it is applied
Gemini Omni Flash Preview375 in · 2,250 text out · 4,375 video outVideo output is pre-authorized at its known worst case
Gemini Omni 1.1 Flash Preview375 in · 2,250 text out · 4,375 video outVideo output is pre-authorized at its known worst case

Questions

Why is there a larger reservation?

The API can choose a video response. Its known maximum is reserved before the call, then unused credits are returned if the result costs less.

Why can an Omni request reserve more credits?

The model may return video, whose known maximum cost is reserved before the request.

Is unused Omni reservation charged?

No. Settlement returns the part not used by the measured response.