Gemini 3 User Guide 2025.12 latest

Google’s most powerful multi-modal AI, trillions of parameters, 1 million tokens context

📋 Quick Overview

📅 Release time

Pro: 2025.11.18 / Flash: 2025.12.17

💰 Pricing

Flash extremely low cost / Pro $2/million tokens

🧠 Parameter scale

Over one trillion parameters, sparse MoE architecture

📝 Context

1 million token (the largest in the industry)

🤖 Gemini 3 model family

🧠 Gemini 3 Deep Think Coming soon

Deep thinking mode, dedicated to complex reasoning tasks

🚀 Gemini 3 core upgrade

🧠 trillion parameters

Sparse MoE architecture, completely based on TPU training

📝 1 million contexts

The largest in the industry, supporting entire code base/website analysis

🎨 Full modal support

Text, image, video, audio, code

🤖Ability of the agent

Antigravity platform, natural language generation of complete code

🔧 Native tools

Google search, code execution, third-party API

🔍 Deep Research

Free and open, automated research report generation

📊 Comparison with competing products (December 2025)

Features Gemini 3 Pro GPT-5.2 Claude Opus 4.5
Context window 1 million tokens ~400K tokens 200K tokens
Maximum output 64K tokens 128K tokens 16K tokens
MMMU-Pro 81% ~75% ~70%
Video Understanding 87.6% Limited ❌
Deep Research ✅ Free ❌ ❌
Enter price $2/million ~$5/million $15/million

🚀 Google Antigravity Agent Platform

The agent software development platform released simultaneously with Gemini 3:

💬 Natural Language Programming

Describe the requirements to generate complete code

🎨 Dynamic interface customization

Adjust UI and interaction logic in real time

🤝 Multi-agent collaboration

Multiple Agents collaborate to perform complex tasks

💰 Pricing Plan

Free version

$0
/month
  • Gemini 3 Flash
  • Deep Research ✅
  • Usage restrictions apply

API Pricing

Model Input Output
Gemini 3 Pro $2/million tokens $12/million tokens
Gemini 3 Flash $0.4/million tokens $2.4/million tokens

❓ FAQ

Gemini 3 Pro vs Flash How to choose?

Pro is suitable for complex tasks that require the strongest performance; Flash is suitable for large-scale deployment and cost-sensitive scenarios, and is extremely cost-effective.

1 million token What is the use of context?

The entire code base, complete technical documentation, and novels can be analyzed at once without segmentation.

Can it be used in China?

Gemini is temporarily unavailable in mainland China and requires an agent to access.