Gemini Enterprise Agent Platform generative AI scale unit (GSU)
Catalog row in Vendor License Metrics · Cited
- Unit
- Other
- Aliases
- GSU; Provisioned Throughput unit
- Definition
- Agent Platform pricing: "Provisioned throughput assures throughput for your generative AI needs and is transacted via generative AI scale units, or GSUs."
- Counting rules
- Size the purchase by converting expected queries per second into burndown-adjusted tokens per second and dividing by the per-GSU throughput for the model, rounding up to the model's minimum GSU purchase increment (the worked example rounds 16.96 to 17 GSUs). Commit terms of 1 week, 1 month, 3 months and 1 year are priced per GSU, with separate Global and non-Global endpoint prices.
- Applies to products
- Gemini Enterprise Agent Platform (formerly Vertex AI)