Vertex AI quotas and limits

Stay organized with collections Save and categorize content based on your preferences.

Number of v5e TPU Chips for serving, per region
Region	Value
us-west1	4
us-west2	Not available
us-west3	Not available
us-west4	Not available
us-central1	Not available
us-east1	Not available
us-east4	Not available
us-east5	Not available
us-south1	Not available
northamerica-northeast1	Not available
northamerica-northeast2	Not available
southamerica-east1	Not available
southamerica-west1	Not available
europe-west2	Not available
europe-west1	Not available
europe-west4	Not available
europe-west6	Not available
europe-west3	Not available
europe-north1europe-central2	Not available
europe-west8	Not available
europe-west9	Not available
europe-southwest1asia-south1	Not available
asia-southeast1	Not available
asia-southeast2	Not available
asia-east2	Not available
asia-east1	Not available
asia-northeast1	Not available
asia-northeast2	Not available
australia-southeast1	Not available
australia-southeast2	Not available
asia-northeast3	Not available
me-west1	Not available
me-central1	Not available
me-central2	Not available
europe-west12	Not available
africa-south1	Not available

Custom-trained model limits

The following limits apply to Vertex AI custom-trained models for a given project and region.

Serving

Limit	Value
Number of replicas per project	200
Number of containers per cluster	25,000

Vertex AI Feature Store

This section lists the quotas and limits for the following:

• Vertex AI Feature Store

• Vertex AI Feature Store (Legacy)

Vertex AI Feature Store

The following quotas apply to a given project and region. For example, in a single project, you can have 200 online serving nodes in us-central1 and another 50 nodes in us-east4.

Quota	Value
Online serving requests per minute	300,000
Maximum number of FeatureOnlineStore instances	10
Maximum number of search requests per minute	6,000,000
Maximum number of online serving nodes across all Optimized FeatureOnlineStore instances in the project	80
Maximum number of FeatureView instances across all FeatureOnlineStore instances	30

Vertex AI Feature Store also has the following limits. You can't request an increase to any of the limits in the following table:

Limit	Value
Maximum number of FeatureGroup resources in a project and location	250
Maximum number of Feature resources within a FeatureGroup	10,000
Maximum size of feature data per entity	5 MB
Number of entity IDs per online serving request (FetchFeatureValues)	1
Maximum length of an entity ID	4076 characters
Storage limit for an Optimized online serving node	200 GB

Vertex AI Feature Store (Legacy)

The following quotas apply to a given project and region. For example, in a single project, you can have 75 concurrent batch jobs in us-central1 and another 75 jobs in europe-west4.

Quota	Value
Online serving requests per minute	300,000
Streaming ingestion requests per minute	60,000
Streaming ingestion write throughput per minute	1.2 GB
Feature creation requests per minute	100
Online serving nodes across all featurestores	30
Concurrent batch jobs (ingestion, serving, and delete feature values combined)	75
Concurrent requests to delete feature values	1
Entity types across all featurestores	75

Vertex AI Feature Store (Legacy) also has the following limits. You can't request an increase to any of the limits in the following table:

Limit	Value
Storage limit for an online serving node	5 TB
Total data in the offline store	Unlimited
Features per entity type	5,000
Number of create, update, and delete featurestore requests per day per project per region	500
For streaming ingestion, the size per request	1 MB
For streaming read, the number of entities that can be included per request	100
For batch import, the number of files that can be included per request	5,000 for Avro or 500 for CSV
For batch serving and exports, the number of features you can request	5,000
For batch ingestion and streaming ingestion, the oldest timestamp for which feature data can be ingested	4,000 days from current date

The data retention limit in Vertex AI Feature Store (Legacy) has the following default value, which you can override:

Data retention defaults	Default value
Data retention in offline store (oldest feature value timestamp after which the values are deleted)	4,000 days from the current date
Data retention in online store (oldest feature value timestamp after which the values are deleted)	4,000 days from the current date

You can override the data retention limit in the following ways:

• To override the data retention limit for the online store, set the online_storage_ttl_days parameter while creating or updating a featurestore.

• To override the data retention limit for the offline store, set the offline_storage_ttl_days parameter while creating or updating an entity type.

Vector Search

The following quotas apply to Vector Search for a given project in each region.

Vertex ML Metadata

The following limits apply to Vertex ML Metadata for a given project in each region.

Vertex AI Pipelines

The following quotas and limits apply to Vertex AI Pipelines for a given project in each region.

^* Pipeline run and task requests beyond this limit are queued until resources are available.

Vertex AI Pipelines have the following limits. Note that, unlike quotas, you can't request an increase to a limit.

^* Pipeline tasks running beyond this limit are cancelled.

Vertex AI Decision Optimization

The following quotas and limits apply to Vertex AI Decision Optimization for a given project in each region.

Colab Enterprise quotas and limits

Colab Enterprise quotas and limits are listed separately. See Colab Enterprise quotas and limits.

Quota increases

If you want to increase any of your quotas for Vertex AI, you can use the Google Cloud console to request a quota increase.

For more information about submitting a quota increase request, see the following sections of Working with quotas:

• Requesting a higher quota

Quotas by region and model

View and edit the quotas in the Google Cloud console

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

1. Go to the Quotas and System Limits page.

Go to Quotas and System Limits

2. To adjust the quota, copy and paste the property aiplatform.googleapis.com/generate_content_requests_per_minute_per_project_per_base_model in the Filter. Press Enter.
3. Click the three dots at the end of the row, and select Edit quota.
4. Enter a new quota value in the pane, and click Submit request.

View the requests per minute (RPM) quotas by region and by model

By default, models 2.0 and later use Dynamic shared quota (DSQ).

Choose a region to view the quotas for each available model:

Increase the quotas

If you want to increase any of your quotas for Generative AI on Vertex AI, you can use the Google Cloud console to request a quota increase.

RAG Engine quotas

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	1,500 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

For more rate limits and quotas, see Generative AI on Vertex AI rate limits.

Batch requests

The quotas and limits for batch prediction requests are the same across all regions.

Concurrent batch prediction request limits

The following table lists the limits for the number of concurrent batch prediction requests:

Limit	Value
Gemini models	8

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

Concurrent batch prediction request quotas

The following table lists the quotas for the number of concurrent batch prediction requests, 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.

Custom-trained model quotas

The following quotas apply to Generative AI on Vertex AI tuned models for a given project and region:

Quota	Value
Restricted image training TPU V3 pod cores per region * supported Region - europe-west4	64
Restricted image training Nvidia A100 80GB GPUs per region * supported Region - us-central1 * supported Region - us-east4	8 2

* Tuning scenarios have accelerator reservations in specific regions. Quotas for tuning are supported and must be requested in specific regions.

Text embedding limits

Each text embedding model request can have up to 250 input texts (generating 1 embedding per input text) and 20,000 tokens per request. Only the first 2,048 tokens in each input text are used to compute the embeddings.

For text-embedding-large-exp-03-07, each request can only include a single input text. The quota for this model is listed under the name text-embedding-large-001.

Gen AI evaluation service service quotas

The Gen AI evaluation service uses gemini-2.0-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 quota is calculated on a per-project basis, which means that any requests directed to gemini-2.0-flash for model inference and model-based evaluation contribute to the quota. 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
Online prediction requests per minute for base_model: gemini-2.0-flash	See Quotas by region and model.

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 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 for a list of models.

Pipeline evaluation quotas

If you receive an error related to quotas while using the evaluation pipelines service, you might need to file a quota increase request. See View and manage quotas for more information. The evaluation pipelines service uses Vertex AI Pipelines to run PipelineJobs. See relevant quotas for Vertex AI Pipelines. The following are general quota recommendations:

Service	Quota	Recommendation
Vertex AI API	Concurrent LLM batch prediction jobs per region	Pointwise: 1 * num_concurrent_pipelines Pairwise: 2 * num_concurrent_pipelines
Vertex AI API	Evaluation requests per minute per region	1000 * num_concurrent_pipelines

Additionally, the relevant quota for model-based metrics depends on which judge model you use:

Tasks	Quota	Base model	Recommendation
summarization question_answering	Online prediction requests per base model per minute per region per base_model	text-bison	60 * num_concurrent_pipelines

Vertex AI Agent Engine

The following quotas and limits apply to Vertex AI Agent Engine for a given project in each region.

Troubleshoot error code 429

To troubleshoot the 429 error, see Error code 429.

What's next

• Learn more about Generative AI on Vertex AI quotas and limits.