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 Pro Ultimate version 2025.11.18
All-round multi-modal flagship, surpassing GPT-5.1 and Claude 4.5
- Context: 1 million token (whole database and site level understanding)
- Output: Up to 64,000 tokens
- Speed: 128 tokens/second
- Multimodal: MMMU-Pro 81%,Video-MMMU 87.6%
- Pricing: Input $2/million, output $12/million
⚡ Gemini 3 Flash Latest 2025.12.17
Pro-level reasoning capability + Flash-level low cost and low latency
- Pricing: Only 1/5 of Claude and 1/4 of GPT
- Coding: Beyond Claude Sonnet 4.5
- Reasoning: Comprehensively crush the same level, competing with GPT-5.2.
- Speed: Optimized for fast response
🧠 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
- Gemini 3 Flash
- Deep Research ✅
- Usage restrictions apply
Gemini Advanced
- Gemini 3 Pro
- 1 million token context
- 2TB Google One
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.