The details:

• Gemini 2.0 Flash performs better than the previous 1.5 Pro model at twice the speed. It supports multimodal inputs (images, video, audio) and can now generate images mixed with text and multilingual text-to-speech output.

• The model integrates natively with tools like Google Search, code execution, and third-party functions. It's available now through the Gemini API in Google AI Studio and Vertex AI, with general availability starting January 2025.

• Google is testing three main research prototypes: Project Astra (a universal AI assistant), Project Mariner (a browser-based task completion tool), and Jules (a coding assistant that works within GitHub workflows).

The tools

• Project Astra: An AI assistant for phones and glasses that supports multiple languages, uses Google Search/Lens/Maps, and maintains conversation memory for up to 10 minutes. The system aims to function as a personal assistant in everyday life.

• Project Mariner: A browser-based AI agent that works through a Chrome extension. It can understand and interact with web content, including text, code, images, and forms to help complete online tasks.

• Google AI Studio: A development platform where developers can access and work with Gemini models through APIs. It's the main access point for testing and implementing Gemini 2.0 Flash.

• Jules: An AI coding assistant that integrates with GitHub workflows.

Why it matters:

Gemini 2.0 represents a major leap in AI capabilities and performance. The release of Gemini 2.0 places Google even closer to OpenAI and Anthropic in the race to have the most capable foundation model and AI chat.