Build with
Gemini 2.0 API
Get Gemini 2.0 Flash API that prioritizes explicit reasoning chains for high performance on complex math and code tasks.
Google
Gemini 2.0 Flash Thinking
Gemini 2.0 Flash Thinking Experimental has improved in transparent AI reasoning, combining accelerated performance with visible thought processes. This specialized model demonstrates 73.3% accuracy on advanced mathematics benchmarks while providing clear insight into its problem-solving methodology.
Cognitive Transparency
Gemini 2.0 Flash Thinking exposes its step-by-step reasoning, allowing users to observe how it arrives at conclusions and verify its logic in real-time.
Extended Memory Architecture
Gemini 2.0 Flash Thinking can handle documents of 1M token capacity, perfect for lengthy academic works and technical manuals.
Smart Tool
Gemini 2.0 Flash Thinking seamlessly integrates code execution and validation capabilities, providing instant feedback on implementation choices.
Google
Gemini 2.0 Flash
Gemini 2.0 Flash Experimental, the first model in the family, is twice as fast as its predecessor, delivering unparalleled performance while optimizing energy use by 67%.
Multimodality
Gemini 2.0 introduces native image generation and controllable text-to-speech capabilities. The model can see, hear, and speak.
Agentic AI
Powered by Google's advanced AI studio and Vertex AI, you can now build AI agents that can navigate, search, and execute tasks on your behalf.
Spatial Understanding 3D
Gemini 2.0's spatial reasoning can interpret and interact with physical spaces, offering applications from home automation to industrial robotics.
New level
Gemini 2.0 Beats GPT o1 preview
Google's Gemini 2.0 is determined to challenge OpenAI's o1 preview dominance, while also competing internally, with 2.0 Flash outperforming 1.5 Pro in key benchmarks at twice the speed.
Versus Gemini 1.5 Flash 002
Gemini 2.0 surpasses Gemini 1.5 Flash in key areas, with significant improvements in MMLU PRO (76.4 vs 67.3). It also excels in math reasoning (89.7 compared to 77.9).
Versus Gemini 1.5 Pro 002
Gemini 2.0 outshines Gemini Pro 1.5 in coding (92.9 vs 85.4) and performs better in reasoning (62.1 vs 59.1) and MMMU (70.7 vs 65.9).
Versus GPT o1 preview
Comparing to o1, Gemini 2.0 excels in handling edge cases during debugging and delivering concise, effective creative outputs. Learn practical test comparison →
AI Agents by Meta’s Gemini 2.0
With Gemini 2.0, Google isn't just updating a model; they're ushering in an era where AI agents become an integral part of our daily lives.
Project Astra
Your new universal assistant that understands your world in real time, providing assistance with memory, navigation, and even remembering past conversations.
Project Mariner
An AI that surfs the web for you, gathering information, filling in forms, and even booking that flight you've been meaning to get to.
Jules
A coding assistant for developers, integrated into your GitHub workflow, reducing the time you spend on bug fixes and mundane tasks.
Real World Applications
Automated Research: Deploy agents that can research, compile, and present data on business trends, competitor analysis, or market conditions.
Project Mariner: Use this prototype to automate and enhance web-based tasks from within Chrome.
Visual Assistance: Help customers troubleshoot issues via video calls, where the AI can see the problem and guide through solutions in real-time.
Language Fluency: Break language barriers with an AI that can switch languages as naturally as humans do, enhancing customer interactions globally.
Dynamic AI Companions: Integrate Gemini agents that can strategize, adapt, and provide real-time advice, enhancing player engagement and experience.
Virtual World Navigation: Assist players in navigating complex virtual worlds with AI that understands the game's 3D space.
Getting started with
Gemini 2.0 Flash API
API Playground to quickly try Gemini 2.0.
For more information about technical features, please refer to the Gemini 2.0 Flash model card here.