This document describes how billing for Gemini-assisted features in Looker will be handled when it goes into effect. Use this page to understand how usage, how token allocations work, and where to find more information.
Usage billing overview
Billing for Gemini-assisted features is tied to the billing account of a Google Cloud project.
- For Looker (Google Cloud core) instances, the Google Cloud project and billing account that were linked with the instance during its provisioning are used for usage billing. You can't change the Google Cloud project that's linked to your Looker (Google Cloud core) instance.
- For Looker (original) instances, a Looker admin must link a Google Cloud project to the instance. To avoid disruption to users, we recommend that you configure this linking before billing takes effect. Although you can link a Looker (original) instance to any Google Cloud project with a valid billing account, you should use the Google Cloud project that you want to be billed for your instance users' Gemini-assisted feature usage.
Once billing takes effect, any usage that extends beyond the instance's included monthly data token allocation will be subject to overage fees. For detailed rates and pricing information, see the Looker pricing page.
Features subject to billing
Key Gemini-assisted features that incur charges include Conversational Analytics and other generally available features that rely on Gemini. You won't be charged for the use of Gemini-assisted features that are in preview.
See the Gemini in Looker overview documentation page for a list of the Gemini-assisted features that are available for your instance.
How usage is measured
Usage of Gemini-assisted features, including Conversational Analytics, is measured in data tokens. Data tokens are the basic units of text and data that are processed by a Large Language Model (LLM). All interactions with Gemini-assisted features are measured in data tokens. These interactions include the questions that are asked, the context that's provided to the model, and the answers that are sent back.
Token usage is categorized into the following types:
- Input data tokens: The information sent to the model to process your request. In Conversational Analytics, input tokens include:
- Natural language questions and prompts that are submitted by users
- Conversation history and context
- LookML model schema metadata (such as view, dimension, and measure definitions, descriptions, and labels) that are used to ground and interpret the request
- Any relevant data context, such as metadata and agent instructions
- Output data tokens: The content generated and returned by the model. In Conversational Analytics, output tokens include:
- Natural language responses, summaries, and suggestions
- API calls or SQL or LookML queries that are generated to retrieve data
- Model reasoning and orchestration steps
- Response visualizations and chart specifications
- Any other text that is generated by the model as part of the response
Data token consumption applies across all Conversational Analytics interfaces, including chats with an Explore, queries with dashboard data agents, custom Explore data agents, embedded agents, and requests made through the Looker API.
Token allocations and overages
To help manage costs, Looker instances include monthly data token allowances:
- Pooled instance allocation: Included monthly data tokens are pooled at the instance level and shared across all active users of that instance. Token usage is not restricted to or capped for individual users.
- User license contributions: The size of your instance's pooled monthly token allocation scales dynamically based on the number and type of active user licenses on your instance (such as standard internal users versus embed users), as well as any baseline allocation included with your Looker instance edition.
- Monthly reset: Included data token allocations reset at the start of each monthly billing cycle and do not roll over from one month to the next.
- Overage billing: If an instance exceeds its included monthly data token allocation, users can continue using Gemini-assisted features without disruption. Additional usage beyond the monthly allocation is billed as overage, measured separately for input data tokens and output data tokens.
For specific included monthly token allowances per user type, baseline instance allocations, and overage rates, see the Looker pricing page.
Monitor Gemini-assisted feature usage
To understand how usage billing may affect your instance, Looker admins can monitor users' Gemini usage and token consumption on the Conversational Analytics System Activity dashboard. The dashboard tracks both input and output token consumption against your instance's monthly allocation to help you monitor usage trends and forecast capacity.