Supported capabilities

Gemini Enterprise Agent Platform provides a suite of advanced features for the Gemini model family. To help you meet specific regulatory and architectural needs, we distinguish between capabilities supported on standard regional endpoints and those supported on jurisdictional multi-region endpoints.

Which models are supported?

Model and capability availability support depends on the endpoint you use:

  • Standard regional endpoints: Supported for all versions of the Gemini family.
  • Jurisdictional multi-region endpoints: These are designed exclusively for the Gemini 3 family and future versions. Older models (like Gemini 2.5 and earlier) are not supported on these endpoints.
  • Global endpoints: Most models are supported on global endpoints. For the complete list of models available on global endpoint, see Deployment and endpoints.

For more information on the different kinds of available endpoints, see Deployment and endpoints.

ML processing commitments

ML processing for Gemini models have the following per-capability commitments:

  • Guaranteed support: The capabilities listed in the following sections are guaranteed to have ML processing support within the specified regional or jurisdictional boundary.
  • Non-Listed capabilities: If a capability is not explicitly listed, ML processing is not guaranteed to occur in a specific location.

For more information about ML processing in Agent Platform, see Data residency.

Endpoint support

The following sections cover the endpoint support for the following model aspects:

Model availability

Model availability covers whether or not a particular model can be used on a specific kind of endpoint. The following table covers which model families are supported for each type of endpoint. For per-model information, see Locations.

For workloads requiring strictly jurisdictional ML processing, you must use the Gemini 3.x family. Models older than Gemini 3 are not supported on jurisdictional multi-region endpoints and will continue to be served using standard regional endpoints.

Model family Standard regional endpoints Jurisdictional multi-region endpoints Global endpoint
Gemini 3.x (Pro and Flash)
Gemini 2 (all versions)

Capabilities, consumption options, and features

The following table covers the availability of different model capabilities across locational endpoints and jurisdictional multi-region endpoints for Gemini 3.x models and higher. If a capability or feature is not explicitly listed as supported, localized ML processing is not guaranteed.

Capability, feature, or consumption option Standard regional endpoints Jurisdictional multi-region endpoints Global endpoint
Capabilities, tools, and features
Chat completions
Code execution
Controllable parameters
Context caching
Count tokens
Function calling
Grounding with Agent Platform Search
Grounding with Google Search
Model tuning
RAG Engine
Structured output
System instructions
Thinking budget
Thought summaries
Consumption options
Batch inference
Standard PayGo (without usage tiers)
Standard PayGo with Usage Tiers
Priority PayGo
Flex PayGo
Provisioned Throughput
Single Zone PT (SZPT)
Platform features
Security controls
Computer Use

For instructions on disabling unsupported capabilities in Gemini Enterprise Agent Platform, see organization policy constraints.

Known constraints

The following features are either out of scope or involve global infrastructure that cannot be localized to a jurisdictional boundary:

  • UI console (Agent Studio): Not supported
  • Gemini Live API: Not supported
  • Evaluation service (SxS): No plan
  • Priority PayGo and Flex PayGo: No plan

What's next