Google Vertex AI Pricing: Complete Enterprise Guide (2026)
Contents
- What is Vertex AI?
- How Vertex AI Pricing Works
- Gemini Model Pricing On Vertex AI (2026)
- Vertex AI Agent Builder And Agent Engine Pricing
- Vertex AI Vs AWS Bedrock Vs Azure AI
- Hidden Costs That Inflate Your Vertex AI Bill
- How To Reduce Vertex AI Costs
- How CloudZero Maps Vertex AI Spend To Business Outcomes
- FAQs
Quick Answer
Google Vertex AI pricing is usage-based with no upfront commitment. The cheapest current model, Gemini 2.5 Flash-Lite, costs $0.10 per million input tokens and $0.40 per million output tokens. The flagship model, Gemini 2.5 Pro, starts at $1.25 per million input tokens and $10.00 per million output tokens. Total monthly costs range from under $100 for prototyping to $100,000+ for enterprise production, depending on model selection, inference volume, training compute, and which Vertex services you stack. Google offers a $300 free trial credit valid for 90 days.
What is Vertex AI?
Vertex AI is Google Cloud’s unified machine learning and generative AI platform. It’s where you train models, deploy them, run inference, build AI agents, and access Google’s Gemini foundation models alongside 200+ third-party models including Anthropic’s Claude, Meta’s Llama, and Mistral, all through a single Vertex AI platform.
As of Google Cloud Next 2026, Vertex AI is transitioning to the Gemini Enterprise Agent Platform, a rebrand consolidating Vertex AI and Agentspace into one product. The billing mechanics haven’t changed. The marketing deck has.
| Category | Services | What it does |
|---|---|---|
| Foundation models | Vertex AI Studio, Vertex AI Model Garden, Vertex AI OpenAI compatibility endpoint | Prompt and test 200+ models (Gemini, Claude, Llama, Mistral). Port existing OpenAI code to Gemini without rewrites |
| AI agents | Vertex AI Agent Builder, Vertex AI Agent Engine | Build and deploy production agents with persistent memory, session management, and managed runtime |
| Search and retrieval | Vertex AI Search, Vertex AI Vector Search, Vertex AI RAG Engine | Enterprise search, similarity search for embeddings, and end-to-end retrieval-augmented generation |
| Media generation | Vertex AI Veo, Imagen, Chirp | Video generation, image generation, and text-to-speech. |
| ML platform | Custom training, AutoML, Vertex AI Workbench, Vertex AI Feature Store, Pipelines | Model training, notebooks, feature management, experiment tracking, and monitoring. |
How Vertex AI Pricing Works
Vertex AI pricing follows Google Cloud’s standard pattern: pay-as-you-go, metered by usage, with tiered pricing discounts at committed volumes. Your bill breaks into four dimensions:
- Generative AI API calls are billed per token, with rates varying by model and context length.
- Agent and application services such as Agent Builder queries, Agent Engine runtime, sessions, memory, and search.
- Training and tuning compute is billed per node-hour for custom training and per training token for supervised fine-tuning.
- Managed infrastructure such as prediction endpoints, vector search nodes, Workbench notebooks, feature store instances, and pipelines are billed per hour.
Gemini Model Pricing On Vertex AI (2026)
Vertex AI Gemini pricing represents the largest share of cost for most teams. Google prices models per million tokens, with separate rates for input and output.
| Model | Input (per 1M tokens) | Output (per 1M tokens) | Context window |
|---|---|---|---|
| Gemini 3.1 Pro | $2.00 (≤200K) / $4.00 (>200K) | $12.00 / $24.00 | 1M tokens |
| Gemini 3.1 Flash-Lite | $0.25 | $1.50 | 1M tokens |
| Gemini 3 Flash | $0.50 | $3.00 | 1M tokens |
| Gemini 2.5 Pro | $1.25 (≤200K) / $2.50 (>200K) | $10.00 / $15.00 | 1M tokens |
| Gemini 2.5 Flash | $0.30 | $2.50 | 1M tokens |
| Gemini 2.5 Flash-Lite | $0.10 | $0.40 | 1M tokens |
Vertex AI Agent Builder And Agent Engine Pricing
Vertex AI Agent Builder pricing covers the full stack for building AI agents, search applications, and conversational interfaces. Vertex AI Agent Engine pricing covers the managed runtime.
| Service | Price | Unit |
|---|---|---|
| Search (data store queries) | $4.00 | Per 1,000 queries |
| Agent Engine runtime | $0.0864 | Per vCPU-hour |
| Sessions and Memory Bank | $0.25 | Per 1,000 events |
Vertex AI Vs AWS Bedrock Vs Azure AI
| Capability tier | Google (Vertex AI) | AWS (Bedrock) | Azure (OpenAI) |
|---|---|---|---|
| Flagship | Gemini 2.5 Pro → $1.25 / $10.00 | Claude 3 Opus → ~$15 / ~$75 | GPT-4o → ~$5 / ~$15 |
| Mid-tier | Gemini 2.5 Flash → $0.30 / $2.50 | Claude 3.5 Sonnet → ~$3 / ~$15 | GPT-4o-mini → ~$0.15 / ~$0.60 |
| Cost-efficient | Gemini 2.5 Flash-Lite → $0.10 / $0.40 | Claude 3 Haiku → ~$0.25 / ~$1.25 | GPT-4o-mini → ~$0.15 / ~$0.60 |
Hidden Costs That Inflate Your Vertex AI Bill
- Idle endpoints. A deployed A100 endpoint costs roughly $2,642/month.
- Thinking token inflation. Reasoning tokens billed as output can double or triple the effective cost of a request.
- Grounding surcharges. Google Search grounding costs can often exceed model inference call costs.
How To Reduce Vertex AI Costs
Seven tactics, ordered by expected savings:
- Route requests to the cheapest adequate model.
- Use the Batch API.
- Cache aggressively.
- Scale endpoints to zero when idle.
- Use committed use discounts for steady GPU workloads.
- Control thinking token spend.
- Get granular cost attribution.
How CloudZero Maps Vertex AI Spend To Business Outcomes
CloudZero connects directly to Google Cloud billing and maps every Vertex AI line item to teams, products, features, and customers.
Cost per customer. Cost per feature. Cost per inference.
CloudZero is relevant to Vertex AI specifically. Instead of a billing line that says "Vertex AI API: $23,847," CloudZero shows what that amount bought in business terms.
FAQs
What is the cheapest Vertex AI model?
Gemini 2.5 Flash-Lite at $0.10 per million input tokens and $0.40 per million output tokens.
Is Vertex AI free to use?
Partially. New Google Cloud accounts get $300 in free credits for 90 days.
Is Vertex AI being deprecated?
No. Vertex AI is actively expanding and transitioning to the Gemini Enterprise Agent Platform.
How does Vertex AI API pricing compare to AWS Bedrock?
Both use pay-per-token models with Vertex AI offering tighter GCP integration.
What are Vertex AI thinking tokens and do they cost money?
Yes, these tokens are billed as output.