This document lists the quotas and system limits that apply to Generative AI
models on Agent Platform.

- **Quotas** have default values, but you can typically request adjustments.
- **System limits** are fixed values that can't be changed.

Google Cloud uses quotas to help ensure fairness and reduce
spikes in resource use and availability. A quota restricts how much of a
Google Cloud resource your Google Cloud project can use. Quotas
apply to a range of resource types, including hardware, software, and network
components. For example, quotas can restrict the number of API calls to a
service, the number of load balancers used concurrently by your project, or the
number of projects that you can create. Quotas protect the community of
Google Cloud users by preventing the overloading of services. Quotas also
help you to manage your own Google Cloud resources.

The Cloud Quotas system does the following:

- Monitors your consumption of Google Cloud products and services
- Restricts your consumption of those resources
- Provides a way to [request changes to the quota value](https://docs.cloud.google.com/docs/quotas/help/request_increase) and [automate quota adjustments](https://docs.cloud.google.com/docs/quotas/quota-adjuster)

In most cases, when you attempt to consume more of a resource than its quota
allows, the system blocks access to the resource, and the task that
you're trying to perform fails.

Quotas generally apply at the Google Cloud project
level. Your use of a resource in one project doesn't affect
your available quota in another project. Within a Google Cloud project, quotas
are shared across all applications and IP addresses.


For more information, see the
[Cloud Quotas overview](https://docs.cloud.google.com/docs/quotas/overview).


There are also *system limits* on Agent Platform resources.
System limits can't be changed.

There are also *limits* on Gemini Enterprise Agent Platform resources. System limits
can't be changed.

## Tuned model quotas

Tuned model inference shares the same quota as the base model.
There is no separate quota for tuned model inference.

## Embedding limits

Requests for `gemini-embedding-001` and `gemini-embedding-2` are subject to global quotas.

| Base model | Quota | Metric |
|---|---|---|
| base_model: gemini-embedding | 100,000,000 | `aiplatform.googleapis.com/global_embed_content_input_tokens_per_minute_per_base_model` |
| base_model: gemini-embedding | 100,000 | `aiplatform.googleapis.com/global_embed_content_requests_per_minute_per_base_model` |
| base_model: gemini-embedding-2 | 200,000,000 | `aiplatform.googleapis.com/global_embed_content_input_tokens_per_minute_per_base_model` |
| base_model: gemini-embedding-2 | 60,000 | `aiplatform.googleapis.com/global_embed_content_requests_per_minute_per_base_model` |

Requests for `gemini-embedding-001` using the `predict` API are also subject to the following quotas:

| Base model | Quota | Metric |
|---|---|---|
| base_model: N/A | 30,000 | `aiplatform.googleapis.com/online_prediction_requests` |

## Agent Runtime quotas

The following quotas apply to [Agent Runtime](https://docs.cloud.google.com/gemini-enterprise-agent-platform/build/runtime) for a given project in each region:

| Description | Quota | Metric |
|---|---|---|
| Create, delete, or update Agent Runtime resources per minute | 10 | `aiplatform.googleapis.com/reasoning_engine_service_write_requests` |
| Create, delete, or update Agent Runtime sessions per minute | 100 | `aiplatform.googleapis.com/session_write_requests` |
| Get, list, or retrieve Agent Runtime sessions per minute | 10000 | `aiplatform.googleapis.com/session_read_requests` |
| `Query` or `StreamQuery` Agent Runtime per minute | 90 | `aiplatform.googleapis.com/reasoning_engine_service_query_requests` |
| Append event to Agent Runtime sessions per minute | 300 | `aiplatform.googleapis.com/session_event_append_requests` |
| Maximum number of Agent Runtime resources | 100 | `aiplatform.googleapis.com/reasoning_engine_service_entities` |
| Create, delete, or update Agent Runtime memory resources per minute | 100 | `aiplatform.googleapis.com/memory_bank_write_requests` |
| Get, list, or retrieve from Agent Runtime Memory Bank per minute | 300 | `aiplatform.googleapis.com/memory_bank_read_requests` |
| Sandbox environment (Code Execution) execute requests per minute | 1000 | `aiplatform.googleapis.com/sandbox_environment_execute_requests` |
| Sandbox environment (Code Execution) entities per region | 1000 | `aiplatform.googleapis.com/sandbox_environment_entities` |
| Sandbox environment (Code Execution) write requests per minute | 500 | `aiplatform.googleapis.com/sandbox_environment_write_requests` |
| A2A Agent post requests like `sendMessage` and `cancelTask` per minute | 60 | `aiplatform.googleapis.com/a2a_agent_post_requests` |
| A2A Agent get requests like `getTask` and `getCard` per minute | 600 | `aiplatform.googleapis.com/a2a_agent_get_requests` |
| Concurrent live bidirectional connections using the `BidiStreamQuery` API per minute | 10 | `aiplatform.googleapis.com/reasoning_engine_service_concurrent_query_requests` |

## Multimodal input quotas

The following quotas apply to multimodal input for `generateContent` and `streamGenerateContent` requests for a given project in each region. Each quota is enforced per base model and resolution. The same limits apply to requests served by the global endpoint, using the corresponding `..._global` metric.

| Description | Quota | Metric |
|---|---|---|
| Generate content requests with image input per minute per base model and resolution | 34,000,000 | `aiplatform.googleapis.com/generate_content_image_input_per_base_model_id_and_resolution` |
| Generate content requests with video input per minute per base model and resolution | 192,000,000 | `aiplatform.googleapis.com/generate_content_video_input_per_base_model_id_and_resolution` |
| Generate content requests with audio input per minute per base model and resolution | 11,000,000 | `aiplatform.googleapis.com/generate_content_audio_input_per_base_model_id_and_resolution` |
| Generate content requests with document input per minute per base model and resolution | 1,200,000 | `aiplatform.googleapis.com/generate_content_document_input_per_base_model_id_and_resolution` |

To request a higher limit for these quotas, contact your Google Cloud account
team; they can't be increased from the Google Cloud console.

## Generative media quotas


To view and edit generative media quotas in the Google Cloud console, do the
following:

1. Make sure that you have [enabled the Agent Platform API](https://docs.cloud.google.com/gemini-enterprise-agent-platform/machine-learning/start/cloud-environment) for your project. If the API isn't enabled, the Agent Platform quotas won't display.
2. Go to the **Quotas and System Limits** page.

   [Go to Quotas and System Limits](https://console.cloud.google.com/iam-admin/quotas/list)
3.
   In the **Filter** text box, enter the following queries:

   `
   Metric: METRIC_NAME


   Dimensions (e.g. location): MODEL_NAME
   `


   Replace the following:
   - <var translate="no">METRIC_NAME</var>: The metric name to search for, one of the following:
     - **Gemini Omni** : `global_generate_content_requests_per_minute_per_project_per_base_model`
     - **Veo** : `long_running_online_prediction_requests_per_base_model`
     - **Lyria** : `global_generate_content_requests_per_minute_per_project_per_base_model`
   - <var translate="no">MODEL_NAME</var>: The name of the model that you are using.
4. To adjust the quota, click **More** and then select **Edit quota**.
5. Enter a new quota value in the pane, and click **Submit request**.

## Batch inference

The quotas and limits for batch inference jobs are the same across all regions.  

### Concurrent batch inference job limits for Gemini models

There are no predefined quota limits on batch inference for Gemini models. Instead, the batch service provides access to a large, shared pool of resources, dynamically allocated based on the model's real-time availability and demand across all customers for that model. When more customers are active and saturated the model's capacity, your batch requests might be queued for capacity.  

### Concurrent batch inference job quotas non-Gemini models

The following table lists the quotas for the number of concurrent batch inference jobs, which don't apply to Gemini models:

| **Quota** | **Value** |
|---|---|
| `aiplatform.googleapis.com/textembedding_gecko_concurrent_batch_prediction_jobs` | 4 |

If the number of tasks submitted exceeds the allocated quota, the tasks are placed in a queue and processed when the quota capacity becomes available.

### View and edit the quotas in the Google Cloud console

> [!CAUTION]
> **Caution:** Before you begin, make sure that you have [enabled
> the Agent Platform API](https://docs.cloud.google.com/gemini-enterprise-agent-platform/machine-learning/start/cloud-environment) for your project. If the API isn't enabled, the Agent Platform quotas won't display.

To view and edit the quotas in the Google Cloud console, do the following:

1. Go to the **Quotas and System Limits** page.
2. [Go to Quotas and System Limits](https://console.cloud.google.com/iam-admin/quotas)
3. To adjust the quota, copy and paste the property `aiplatform.googleapis.com/textembedding_gecko_concurrent_batch_prediction_jobs` in the **Filter** . Press **Enter**.
4. Click the three dots at the end of the row, and select **Edit quota**.
5. Enter a new quota value in the pane, and click **Submit request**.

## Semantic Governance Policy quotas

The following quotas apply to [Semantic Governance Policy](https://docs.cloud.google.com/gemini-enterprise-agent-platform/govern/policies/semantic-governance-overview) for a given project in each region:

| Description | Quota | Metric |
|---|---|---|
| Get or list Semantic governance policy resources per minute | 600 | `aiplatform.googleapis.com/semantic_governance/policy_read_requests` |
| Create, update, or delete Semantic governance policy resources per minute | 60 | `aiplatform.googleapis.com/semantic_governance/policy_write_requests` |
| Get Semantic governance policy engine resources per minute | 600 | `aiplatform.googleapis.com/semantic_governance/engine_read_requests` |
| Update or deprovision Semantic governance policy engine resources per minute | 60 | `aiplatform.googleapis.com/semantic_governance/engine_write_requests` |
| Number of Semantic governance policy resources per project per location. | 1,000 | `aiplatform.googleapis.com/semantic_governance/policy_count` |

### View and edit the quotas in the Google Cloud console

> [!CAUTION]
> **Caution:** Before you begin, make sure that you have [enabled
> the Agent Platform API](https://docs.cloud.google.com/gemini-enterprise-agent-platform/machine-learning/start/cloud-environment) for your project. If the API isn't enabled, the Agent Platform quotas won't display.

To view and edit the quotas in the Google Cloud console, do the following:

1. Go to the **Quotas and System Limits** page.
2. [Go to Quotas and System Limits](https://console.cloud.google.com/iam-admin/quotas)
3. To adjust the quota, copy and paste the property `aiplatform.googleapis.com/semantic_governance/policy_write_requests` in the **Filter** . Press **Enter**.
4. Click the three dots at the end of the row, and select **Edit quota**.
5. Enter a new quota value in the pane, and click **Submit request**.

## RAG Engine on Gemini Enterprise Agent Platform

The [VPC-SC security controls](https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/security-controls) and
CMEK are supported by Agent Platform RAG Engine. Data residency and AXT security controls aren't
supported.
For each service to perform retrieval-augmented generation (RAG) using RAG Engine, the following quotas apply, with the quota measured as requests per minute (RPM).

| Service | Quota | Metric |
|---|---|---|
| RAG Engine data management APIs | 60 RPM | `VertexRagDataService requests per minute per region` |
| `RetrievalContexts` API | 600 RPM | `VertexRagService retrieve requests per minute per region` |
| `base_model: textembedding-gecko` | 1,500 RPM | `Online prediction requests per base model per minute per region per base_model` An additional filter for you to specify is `base_model: textembedding-gecko` |

The following limits apply:

| Service | Limit | Metric |
|---|---|---|
| Concurrent `ImportRagFiles` requests | 3 RPM | `VertexRagService concurrent import requests per region` |
| Maximum number of files per `ImportRagFiles` request | 10,000 | `VertexRagService import rag files requests per region` |

## Gen AI evaluation service

The Gen AI evaluation service uses Gemini 2.5 Flash as a default judge model for model-based metrics. A single evaluation request for a model-based metric might result in multiple underlying requests to the Gen AI evaluation service. Each model's consumption is calculated at the organization level, which means that any requests directed to judge model for model inference and model-based evaluation contribute to the model's consumption. Quotas for the Gen AI evaluation service and the underlying judge model are shown in the following table:

| **Request quota** | **Default quota** |
|---|---|
| Gen AI evaluation service requests per minute | 1,000 requests per project per region |
| Gemini throughput | Depends on model and [consumption option](https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/standard-paygo) |
| Concurrent evaluation runs | 20 concurrent evaluation runs per project per region |

If you receive an error related to quotas while using the Gen AI evaluation service, you might
need to file a quota increase request. See [View and manage
quotas](https://docs.cloud.google.com/docs/quotas/view-manage) for more information.

| **Limit** | **Value** |
|---|---|
| Gen AI evaluation service request timeout | 60 seconds |

When you use the Gen AI evaluation service for the first time in a new project, you might
experience an initial setup delay up to two minutes. If your first request fails, wait a few minutes
and then retry. Subsequent evaluation requests typically complete within 60 seconds.

The maximum input and output tokens for model-based metrics depend on the model used
as the judge model. See [Google models](https://docs.cloud.google.com/gemini-enterprise-agent-platform/models) for a
list of models.

### Gemini Enterprise Agent Platform Pipelines quotas

Each tuning job uses Gemini Enterprise Agent Platform Pipelines. For more information,
see [Agent Platform Pipelines quotas and limits](https://docs.cloud.google.com/gemini-enterprise-agent-platform/machine-learning/quotas#vertex-ai-pipelines).

## What's next

Overview

### [Standard PayGo](https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/standard-paygo)


Learn about Standard PayGo, a Agent Platform consumption option lets you pay only for the resources that you consume, without requiring upfront financial commitments.

Resource

### [Agent Platform quotas and system limits](https://docs.cloud.google.com/gemini-enterprise-agent-platform/machine-learning/quotas)


Quotas and system limits related to Agent Platform, excluding product-specific quotas and system limits.

Overview

### [Google Cloud quotas](https://docs.cloud.google.com/docs/quotas/overview)


Learn about how Google Cloud restricts how much of a resource your Google Cloud project can use, and how quotas apply to a range of resource types, including hardware, software, and network components.