Single Zone Provisioned Throughput lets you reservethroughput in specific regions where only onezone isavailable. This option providespredictable performance for Gemini models in use cases where MLprocessing is required.

To view the list of supported models and regions, seeDeployments and endpoints. For the list ofregions and models that support ML processing, seeML processing.

Features of Single Zone Provisioned Throughput

This section outlines the key features of Single Zone Provisioned Throughput:

• Pricing and units are consistent with standard Provisioned Throughput:Single Zone Provisioned Throughput uses the same measure of throughput (GSUs),pricing, and terms asstandardProvisioned Throughput.

• Single Zone Provisioned Throughput supports in-region ML processing: All requests are processed in thepurchased region, including traffic that exceeds your purchased amount ofthroughput. This traffic is billed at thepay-as-you-go rateusing buffer capacity in the region.

• You control the overages: You cancontrol overflow trafficusing the same headers as with standard Provisioned Throughput.

• You can monitor your order: You can monitor your Single Zone Provisioned Throughput order using the existingProvisioned Throughput monitoring capabilities.

Limitations

Single Zone Provisioned Throughput has the following limitations:

• Single Zone Provisioned Throughput is not a Covered Service and is excluded from theGemini Online Inference on Vertex AI Service Level Agreement.

• Single Zone Provisioned Throughput does not integrate with or supportBatch requestsorFine Tuning.

• In regions without ML processing, latency for Single Zone Provisioned Throughput might be higher thanstandard Provisioned Throughput or pay-as-you-go.

Purchase Single Zone Provisioned Throughput

For assistance with purchasing Single Zone Provisioned Throughput,contact your Google Cloud account representative.

What's next

• Purchase standard Provisioned Throughput.