Prices are listed in US Dollars (USD).

Pricing for listed Generally Available (GA) functionality is invoked on Sep 1, 2023.

Vertex AI Search lets developers, even those with limited machine learning skills, tap into the power of Google's foundation large language models, search and recommendations expertise to create enterprise-grade generative AI applications

✩ Google-internal note: For more information on pricing, see go/vertexsc-pricing.

Vertex AI Search provides the ability to quickly build Search Engines for website, unstructured data, structured data to retrieve information and generate grounded answers. You can read more about the features available in Vertex AI Search.

Vertex AI Search offers two pricing models: General and Configurable. You must choose one model for your search application and its associated data stores.

Pricing for Vertex AI Search Index Data Storage

Vertex AI Search for Healthcare provides the ability to quickly build medically tuned Search Engines over healthcare data.

Vertex AI Search for Healthcare includes some features in Preview such as GenAI answers, streaming updates to the index, and others. These features may be priced differently than the current listed price when they are released to General Availability.

Pricing Example for Healthcare Search

Assumption:

• 1,000,000 healthcare search requests a month
• 1,000 GiB of healthcare data indexed

Vertex AI Search for Media enables you to provide highly relevant video results, leveraging Google's query and contextual understanding to improve discovery across your media site.

The only Media Recommendations operations that incur charges are training, tuning, or requesting predictions by calling the recommend method. There's no charge for importing or managing user events or document information.

Training (per node per hour) costs are charged on a daily basis if your model is actively training or if you have submitted a request to resume training. After you pause or delete a model, you are no longer charged. See the documentation for managing training.

Tuning (per node per hour) costs for active models are charged after the tune completes successfully. You are only charged for an incomplete tune if you pause or delete a model during an ongoing tune. In this case, you are then charged for the node hours that were consumed before the model tuning stopped. See the documentation for managing tuning.

Assumption:

• 1,000,000,000 prediction requests a month
• Trains a single model per day, which automatically retrains once per day
• Amounts to about 500 node hours of model training and 100 hours of model tuning per month

Media Recommendations logs an error to Google Cloud Observability for each API request that results in an error, such as a user event request that contains malformed JSON, or a document import request with a negative price. Media Recommendations also logs an error for every prediction request with a document that is not in the imported datastore.

Google Cloud Observability charges by the GiB of logs stored and for retention beyond the default retention period. For details about the free allotment and data retention, see the Google Cloud Observability pricing page.

The size of the logging data depends on the size of your JSON payload, but a GiB would be approximately 200,000 Media Recommendations errors.

The Grounded Generation API enables you to create generative answers to your prompts using information on Google Search or your own data.

The user uses Vertex AI Search and additional grounding facts to generate grounded answers. Each input prompt is 2,500 characters (including inline grounding facts) and each output prompt 200 characters long. The user has selected Gemini 1.5 Flash.

The user uses Google Search to generate grounded answers. Each input prompt is 500 characters and each output prompt 200 characters long. The user has selected Gemini 1.5 Flash.

Check grounding provides the ability to determine how grounded a piece of text (the answer candidate) is in a given set of reference texts (the facts).

For full pricing information of all Document AI features, refer to the Document AI pricing page.

For the Document AI features integrated with and billed through Vertex AI Search, refer to the tables below.

*The size of a page depends on the file format.:

• Images (JPEG/JPG, PNG, BMP, HEIF): Each image = 1 page
• PDF: Each page in the PDF = 1 page
• TIFF: Each image in the TIFF = 1 page
• Word (DOCX): Up to 3,000 characters = 1 page
• Excel (XLSX): Each tab = 1 page
• Powerpoint (PPTX): Each slide = 1 page
• HTML: Up to 3,000 characters = 1 page
• Parsed Documents: Up to 3,000 characters = 1 page

The ranking API takes a list of documents and reranks those documents based on how relevant the documents are to a query.

Compared to embeddings, which look only at the semantic similarity of a document and a query, the ranking API can give you precise scores for how well a document answers a given query.

The ranking API can be used to improve the quality of search results after retrieving an initial set of candidate documents.

A query is defined as having up to 100 documents, though a user can specify more than 100 documents per query. In the case where more than 100 documents are specified, pricing increases by 1 for every multiple of 100 documents.

• For example:
• 132 documents to rank = 2 queries
• 200 documents to rank = 2 queries
• 399 documents to rank = 4 queries
• 401 documents to rank = 5 queries

• Read the Vertex AI Search documentation.
• Try the Pricing calculator.