specification: FinOps Framework
specificationVersion: '1.0'
schema: https://www.finops.org/framework/
alignedWith:
  framework: FinOps Foundation Framework
  frameworkUrl: https://www.finops.org/framework/
  dataSpec: FOCUS
  dataSpecVersion: '1.3'
  dataSpecUrl: https://focus.finops.org/focus-specification/v1-3/
provider: Google Gemini
providerId: google-gemini
publisherName: Google LLC
serviceCategory: AI Infrastructure / LLM
created: '2026-05-04'
# Provenance stamped 2026-08-11: this artifact was written by the API Evangelist
# bulk sweep dated 2026-05-04, not harvested from the provider. See roadmap#35.
method: generated
modified: '2026-05-05'
reconciled: true
tags:
  - Generative AI
  - LLM
  - AI Infrastructure
  - Google
  - FinOps
  - FOCUS
description: 'FOCUS-aligned FinOps for Gemini API: per-model token-based pay-as-you-go pricing with separate
  input vs output rates and multimodal-vs-audio bands; long-context price tiers above 200K-token prompts;
  Batch API offering 50% off; optional Google Search grounding meter; Enterprise via Vertex AI with provisioned
  throughput. Spend rolls into the standard Google Cloud Billing pipeline.'
sources:
  - https://ai.google.dev/pricing
  - https://ai.google.dev/gemini-api/docs/rate-limits
  - https://cloud.google.com/vertex-ai/generative-ai/pricing
billingModel:
  pricingCategory: Token-Based Usage
  billingFrequency: Monthly
  billingCurrency: USD
  chargeCategories:
    - Usage
    - Purchase
    - Tax
    - Adjustment
    - Credit
focusColumns:
  ServiceName: Gemini API
  ServiceCategory: AI Infrastructure
  ProviderName: Google Cloud
  PublisherName: Google LLC
  InvoiceIssuerName: Google LLC
  BillingCurrency: USD
  ChargeCategory: Usage
meters:
  - name: tokens_input
    description: Input tokens billed per model and modality band
    unit: token
    aggregation: sum
    dimensions:
      - model
      - modality
      - prompt_length_band
      - project
  - name: tokens_output
    description: Output tokens billed per model and band
    unit: token
    aggregation: sum
    dimensions:
      - model
      - prompt_length_band
      - project
  - name: tokens_cached
    description: Tokens served from context cache (lower rate than fresh input)
    unit: token
    aggregation: sum
    dimensions:
      - model
      - cache_id
      - project
  - name: batch_tokens_input
    description: Input tokens billed via the Batch API (50% discount)
    unit: token
    aggregation: sum
    dimensions:
      - model
      - project
  - name: batch_tokens_output
    description: Output tokens billed via the Batch API (50% discount)
    unit: token
    aggregation: sum
    dimensions:
      - model
      - project
  - name: search_grounding_queries
    description: Google Search grounding queries (free up to 5K/month, then $14/1K)
    unit: request
    aggregation: sum
    dimensions:
      - model
      - project
  - name: provisioned_throughput
    description: Reserved Gemini capacity hours via Vertex AI Enterprise
    unit: hour
    aggregation: sum
    dimensions:
      - model
      - region
      - project
apis:
  - name: Gemini API (AI Studio)
    baseURL: https://generativelanguage.googleapis.com
    serviceName: Gemini API
    serviceCategory: AI Infrastructure
  - name: Gemini via Vertex AI
    baseURL: https://aiplatform.googleapis.com
    serviceName: Vertex AI Generative AI
    serviceCategory: AI Infrastructure
principles:
  - name: Visibility
    description: Enable Cloud Billing detailed export to BigQuery filtered to the Gemini API SKU; surface
      per-call token counts via the response usage metadata and your application logs for per-feature
      attribution.
  - name: Allocation
    description: Tag Gemini calls with team/feature labels; use separate API keys (or Cloud projects)
      per team for first-class cost allocation; for Vertex AI usage, tag the Vertex AI endpoint with org
      labels.
  - name: Optimization
    description: Choose the smallest model that meets quality (Flash-Lite for routine traffic, Flash for
      mixed workloads, Pro only for high-stakes prompts); enable context caching for repeated long-system-prompt
      patterns; route async/non-latency-sensitive workloads through the Batch API for 50% discount; trim
      prompts to stay under the 200K-token long-context band; share the 5K free Search grounding allowance
      across teams.
  - name: Accountability
    description: Set per-project Cloud Billing budgets aligned with Gemini API tier caps ($250/$2,000/$20,000+);
      designate a model owner per feature; review monthly token burn vs business value (cost-per-resolved-ticket,
      cost-per-classification, etc.).
maintainers:
  - FN: Kin Lane
    email: kin@apievangelist.com