The Gemini Deep Research API represents a major step forward in agentic AI workflows within the Google AI Gemini API ecosystem. By combining autonomous planning, web search, and long-form report generation, developers can build systems that perform deep research tasks automatically.
For teams working on AI automation, data intelligence, or research platforms, integrating the Gemini AI API key through the Gemini API Cloud Console enables powerful research capabilities that previously required hours of manual work.
As the API evolves beyond preview, we expect expanded tools, better limits, and possibly a Gemini Deep Research API free tier in future releases.
In this guide, we explain Gemini Deep Research API documentation, pricing, limits, integration examples, and how to start using the Gemini API key in real developer workflows.
Quick Summary Table (For Busy Developers)
| Feature | Details |
|---|---|
| API Name | Gemini Deep Research API |
| Platform | Google AI Gemini API |
| Model | Gemini 3.1 Pro |
| Agent ID | deep-research-pro-preview-12-2025 |
| Access Method | Gemini API via Google AI Studio or Gemini API Cloud Console |
| Authentication | Gemini AI API Key |
| API Type | Agentic multi-step research API |
| Task Duration | Up to 60 minutes per research task |
| Pricing | $2–$4 per 1M input tokens / $12–$18 per 1M output tokens |
| Tool Cost | $14 per 1,000 Google Search queries (after free allowance) |
| Free Tier | Not available for Deep Research preview |
| Typical Cost | $2–$5 per research task |
| Main Use Cases | Market research, AI agents, automation, competitive analysis |
| Official Docs | Gemini Deep Research API documentation |
You should need to know, Gemini 3 Flash API – Docs, Setup & API Key | Developers Need to Know
What Is Gemini Deep Research API?
The Gemini Deep Research API is an agentic research system powered by Gemini 3.1 Pro within the Google AI Gemini API platform.
Instead of responding to a simple prompt, the system behaves like a research agent that:
- Plans a research strategy
- Searches the web automatically
- Reads and analyzes multiple sources
- Iterates across findings
- Produces structured reports with citations
- Generates tables and summarized insights
This capability is especially useful for developers building tools for:
- Market research automation
- Competitive analysis
- Financial intelligence platforms
- Startup research tools
- Knowledge automation systems
- AI agents and copilots
The API works through the Gemini Interactions API, allowing asynchronous research tasks that can run in the background.
How Gemini Deep Research API Works?
When a request is sent to the Google AI Gemini API, the Deep Research agent performs a multi-step workflow:
- Planning Phase
The AI agent analyzes the prompt and generates a research plan. - Search Phase
It automatically uses Google Search tools to gather relevant sources. - Reading & Extraction
The agent reads web content and extracts structured data. - Analysis & Iteration
It may run multiple research passes to refine findings. - Report Generation
The final output includes detailed explanations, tables, and references.
Because the system performs multiple reasoning steps, tasks may take several minutes rather than seconds.
Gemini Deep Research API Documentation
The official Gemini Deep Research API documentation explains how developers can integrate the research agent using the Gemini API URL and authentication via a Gemini AI API key.
Developers can access the API through:
- Google AI Studio
- Gemini API Cloud Console
- Direct API requests via REST, Python, or JavaScript
Official docs:
https://ai.google.dev/gemini-api/docs/deep-research
Required Parameters
To trigger the research agent, developers must specify:
agent='deep-research-pro-preview-12-2025'background=true
The background mode allows the request to run asynchronously while the client polls for results.
How to Get a Gemini AI API Key?
Before using the Gemini Google API, developers must generate a Gemini AI API key.
Steps
- Open the Gemini API Cloud Console
- Create or select a Google Cloud project
- Enable Google AI Gemini API
- Generate a Gemini API key
- Store the key securely in environment variables
Example:
export GEMINI_API_KEY="your_api_key_here"
This key authenticates requests sent to the Gemini API URL.
Gemini Deep Research API Integration Example
Below is a basic Python example showing how developers can run a research task using the Gemini Deep Research API agent.
from google import genaiclient = genai.Client()interaction = client.interactions.create(
input="Research the EV battery industry landscape and include a table comparing major manufacturers.",
agent='deep-research-pro-preview-12-2025',
background=True
)print(interaction)
Because the research agent runs asynchronously, your application should poll the interaction status until completion.
Developers can also implement:
- streaming responses
- reconnection for long tasks
- structured output prompts
Gemini Deep Research API Pricing
The Gemini AI Gemini API pricing for Deep Research follows the standard Gemini 3.1 Pro token model plus additional tool costs.
Token Pricing
| Tier | Input Tokens | Output Tokens | Free Tier |
|---|---|---|---|
| Paid Standard | $2.00 / 1M tokens | $12.00 / 1M tokens | No |
For larger context windows (over 200K tokens):
| Tier | Input | Output |
|---|---|---|
| Extended Context | $4.00 / 1M tokens | $18.00 / 1M tokens |
Tool Usage Costs
Search tool usage may incur extra charges.
Example:
- $14 per 1,000 search queries after the free allowance.
Typical Cost per Task
| Task Type | Estimated Cost |
|---|---|
| Moderate research | $2 – $3 |
| Complex multi-source research | $3 – $5 |
At the moment, Gemini Deep Research API free tier is not available, since the model is currently in preview.
Gemini Deep Research API Limits
The Gemini Deep Research API limit depends on project quotas configured in Google Cloud.
Key Limits
Maximum task duration
- Up to 60 minutes per research task
Rate limits
- Requests per minute
- Tokens per minute
- Daily request quotas
Quota reset
- Reset occurs daily at midnight Pacific Time (PT).
Since the feature is currently in beta preview, these limits may change as the platform evolves.
Supported Features
The Gemini AI Deep Research API agent currently supports several built-in capabilities.
Available Features
- Autonomous research planning
- Web search integration
- Long-form report generation
- Structured tables
- Source citations
- Asynchronous execution
- File Search support for developer data
Current Limitations
- Custom tools not yet supported
- Audio input not supported
- Preview model subject to updates
Real Developer Use Cases
From our internal development testing while building AI automation systems, the Google AI Gemini API with Deep Research performs especially well for:
1. Market Research Automation
Generating structured reports about industries or competitors.
2. Investment Research Tools
Analyzing startups, financial sectors, or emerging technologies.
3. AI Copilots for Analysts
Allowing analysts to trigger deep research tasks automatically.
4. Knowledge Platforms
Building research agents for SaaS knowledge systems.
Because the Gemini Deep Research agent can read and synthesize many sources automatically, it significantly reduces manual research time.
Best Practices for Developers
To get the best results when using the Gemini AI API key, developers should structure prompts carefully.
Recommended Prompt Structure
Ask the agent to include:
- Executive summary
- Structured sections
- Data tables
- Sources or references
Example prompt:
Research the global EV battery industry.Include:
- Executive summary
- Market size analysis
- Major manufacturers comparison table
- Key trends
Structured prompts produce much better outputs.
Gemini API Free Alternatives
While the Gemini Deep Research API free tier is unavailable, developers can still experiment with:
- Standard Gemini API free tier
- Gemini models through Google AI Studio
- Limited free usage in development environments
However, advanced agent-based research tasks require the paid Gemini 3.1 Pro tier.
Learn free, Gemini API vs OpenAI (I Tested Both) – Which One Should You Actually Use in 2026?
FAQs
What is Gemini Deep Research API?
The Gemini Deep Research API is an agent-based capability in the Google AI Gemini API that performs multi-step research tasks automatically. It can plan research, search the web, analyze content, and generate detailed reports with citations using the Gemini 3.1 Pro model.
How do I get a Gemini AI API key?
You can obtain a Gemini AI API key by opening the Gemini API Cloud Console, creating a Google Cloud project, enabling the Google AI Gemini API, and generating a secure API key for authentication.
Is Gemini Deep Research API free?
No, the Gemini Deep Research API free tier is currently unavailable because the model is still in preview. Developers must use the paid Gemini 3.1 Pro pricing tier to access this feature.
What are the limits of Gemini Deep Research API?
The Gemini Deep Research API limit allows tasks to run for up to 60 minutes. Additional limits include requests per minute, tokens per minute, and daily project quotas managed through the Gemini API Cloud Console.
How much does Gemini Deep Research API cost?
The Gemini AI Gemini API pricing typically costs $2–$4 per 1 million input tokens and $12–$18 per 1 million output tokens, plus additional fees for tools like Google Search queries.
What can developers build with Gemini Deep Research API?
Developers can use the Gemini Google API to build applications like automated research assistants, market analysis tools, financial intelligence platforms, AI copilots, and knowledge automation systems.