The Agent Platform provides access to broad classes of large language models (LLMs) and generative AI models, including Google first-party models, partner models, and open-weights models. This page provides an index of documentation for these models, organized by model types, access options, capabilities, customization, evaluation, and safety.
Get started
Get started with using models on the Agent Platform.
Quickstart
Quickly start calling APIs on the platform.
Get an API key
Learn how to create and manage API keys.
Setup Authentication
Configure application default credentials.
Google GenAI Libraries
Download and install libraries for major programming languages.
Model Garden
Model Garden provides a centralized location to discover, test, and deploy a wide variety of first-party, partner, and open-source models.
Overview of Model Garden
Discover the models available in the catalog.
Use Models in Model Garden
Learn how to interact with and consume models.
Quickstart
Test model capabilities quickly.
Deploy Open Models
Instructions for deploying open models from to endpoints.
Model Types
Agent Platform provides access to a wide variety of models, including:
Google Models
Flagship models developed by Google, including Gemini, Gemma, Imagen, Veo, and Lyria.
Partner Models
Models from leading AI providers integrated into the platform, such as Anthropic Claude, Grok, and Mistral AI.
Open Models
Support for a wide range of open-weights models like DeepSeek, Llama, and Qwen.
Model Access and Deployment
Depending on the model and your requirements, you can access models in two ways:
Model as a Service (MaaS)
Use managed APIs for partner and open models without managing infrastructure.
Model Garden
Discover, test, and customize a wide variety of first-party, partner, and open-source models.
Model Capabilities
Explore advanced capabilities supported by the models:
Long Context
Understand how to work with models that support large context windows.
Thinking Models
Learn about models with reasoning and thinking capabilities.
Computer Use
Documentation on models capable of interacting with computer interfaces.
Bounding Box Detection
Detect objects and return coordinates.
Model Customization (Tuning)
Tailor models to your specific needs:
Tune Models Overview
An introduction to model tuning on Agent Platform.
Gemini Supervised Tuning
Fine-tune Gemini models using supervised training.
Gemini Preference Tuning
Align model behavior using human feedback or preference data.
Tune Embeddings
Customize text or multimodal embeddings.
Model Evaluation
Assess model performance objectively:
Evaluation Overview
Learn about the Gen AI evaluation service, including adaptive rubrics.
Run Evaluation
How to execute evaluation jobs.
Evaluation using Console
Guided workflow via the web UI.
Evaluation using SDK
Programmatic evaluation using Python.
Safety & Security
Build responsible AI applications:
Safety Overview
Overview of safety features and content filtering.
Responsible AI
Guidelines and tools for responsible AI usage.
Abuse Monitoring
How abuse monitoring is handled.
Consumption options
Understand the available consumption options:
Flex Pay-as-you-go
Learn about the Flex pricing model.
Priority Pay-as-you-go
Learn about the Priority pricing model.
Standard Pay-as-you-go
Learn about the Standard pricing model.
