modelInferenceUnitCount property

int modelInferenceUnitCount
final

Optional. The number of model inference units to use for this deployment. This can only be specified for DEPLOYMENT_TYPE_PROD. The following table lists the number of model inference units for different model types:

  • Gemini 2.5 Flash
    • Foundation FMIU: 25
    • Expansion FMIU: 4
  • Gemini 2.5 Pro
    • Foundation FMIU: 32
    • Expansion FMIU: 16
  • Veo 3.0 (undistilled)
    • Foundation FMIU: 63
    • Expansion FMIU: 7
  • Veo 3.0 (distilled)
    • Foundation FMIU: 30
    • Expansion FMIU: 10

Implementation

final int modelInferenceUnitCount;