All guides
Practice now
D3 · AI and innovation with Google Cloud

Generative AI with Google Cloud

Gemini, Vertex AI Studio, Model Garden, grounding, and enterprise generative AI. Everything you need for the CDL exam.

Gemini: Google's foundation model

is Google's family of foundation models. "Multimodal" means it can understand and generate multiple types of content: text, images, code, audio, and video — all in the same model.

It is the primary model on Vertex AI, now part of Gemini Enterprise Agent Platform (the current generation is Gemini 3.1).

ModelDescriptionBest for
Gemini's enterprise offering, integrated into Google Cloud and Workspace. Lets employees discover, build, run, and govern AI agents with corporate security and privacy controls.Advanced reasoning, analysis of long documents, integration with corporate data
Gemini ProBalance between capability and speed. Excellent for most enterprise tasks.Chatbots, summaries, information extraction, code generation
Gemini FlashFast and efficient. Optimized for low latency and reduced cost.High-concurrency applications, classification, repetitive tasks
Gemini NanoDesigned to run on devices (edge/mobile). No cloud connection required.On-device features on Android Pixel phones, offline mobile apps

Vertex AI for Generative AI

Vertex AI Studio

A graphical interface for experimenting with generative models. Test prompts, adjust parameters (temperature, top-p, top-k), and compare responses from different models without writing code.

Test Gemini, Imagen, embeddings, and more
Prompt Gallery with industry-specific examples
Model evaluation and comparison
Automatically generates integration code

Model Garden

A unified catalog of more than 200 foundation models on Vertex AI: Google's models (Gemini, Imagen, Veo, Chirp) and partner models such as Anthropic's Claude and Meta's Llama. It is the central entry point for exploring, comparing, and deploying any model on GCP.

200+ models from Google and partners
Google models: Gemini, , Veo, Chirp, Lyria
Third-party and open: Claude, Llama, Mistral,
One-click deployment to managed endpoints

Grounding and RAG

Connecting the model to real data sources to reduce hallucinations and provide up-to-date information.

with Google Search: responses with current data from the internet
Grounding with Vertex AI Search: searches your own documents
: Retrieval-Augmented Generation with and vector search
Citations: the model cites the sources of its responses

Gemini for Google Workspace

Gemini is integrated directly into Google Workspace applications for businesses. No need to switch tools — the AI shows up where employees already work.

Gmail

Summarize email threads, draft replies, translate messages instantly.

Google Docs

Generate drafts, rephrase paragraphs, change tone (formal/informal), summarize documents.

Google Sheets

Generate formulas with natural language, analyze data, automatically create charts.

Google Meet

Take notes automatically, summarize meetings, translate in real time.

Gemini Enterprise Agent Platform — building AI agents

(the evolution of Vertex AI and Vertex AI Agent Builder since April 2026) is Google Cloud's unified platform for building, deploying, and managing AI agents. It unifies ML models, generative AI, and agent-building tools in a single environment.

For the CDL exam: when the scenario mentions "autonomous agents," "orchestrating tasks with AI," or "connecting an LLM with your own tools and data" → Gemini Enterprise Agent Platform.

Conversational agents

Build chatbots and virtual assistants that can query APIs, databases, and execute actions — not just respond with text.

Data agents

Agents that access BigQuery, Cloud Storage, and other data sources to answer analytical questions in natural language.

Multi-agent orchestration

Coordinates multiple specialized agents that collaborate to complete complex, multi-step tasks.

Responsible AI at Google Cloud

Fairness

Models must perform well for all groups of people without discrimination. Google uses tools like the What-If Tool to detect bias.

Interpretability

Models must be able to explain their predictions. Vertex AI Explainable AI provides feature attributions.

Privacy

Training data is not used to train Google's models. Differential Privacy available for sensitive data.

Safety

Vertex AI has content filters to prevent harmful outputs. Safety settings configurable by category.

Reliability

Models must be reliable and consistent. Continuous evaluation and drift monitoring in production.

Did you get this topic?

Test what you just learned

A company wants to build an internal chatbot to answer questions about its HR policies and documents. The chatbot must answer only using information from its official documents and not make up answers. Which Vertex AI solution is most suitable?

Ready to test "Generative AI with Google Cloud"?

Put what you learned to the test with original CDL exam questions.

Start CDL mock exam

Unlimited mock exams with the Pro plan

Log in to track your progress.