SaMedia, Dec11—Google has unveiled Gemini 2.0, its most advanced AI model, promising to revolutionize AI with enhanced multimodal capabilities and agentic functions. This release builds on the success of Gemini 1.0 and 1.5, which introduced natively multimodal AI, enabling the processing of text, video, images, audio, and code.
“With Gemini 2.0, we’re not just pushing the boundaries of AI; we’re reshaping them,” said Sundar Pichai, CEO of Google and Alphabet. The new model can think, plan, and take actions, bringing the vision of a universal assistant closer to reality.
“We’re getting 2.0 into the hands of developers and trusted testers today. And we’re working quickly to get it into our products, leading with Gemini and Search. Starting today our Gemini 2.0 Flash experimental model will be available to all Gemini users.” Pichai noted.
Gemini 2.0 Flash, the workhorse model, offers low latency and enhanced performance. It supports multimodal inputs and outputs, including natively generated images and steerable text-to-speech in multiple languages. The model also integrates with tools like Google Search, code execution, and third-party functions. Available now to developers via Google AI Studio and Vertex AI, general availability will follow in January.
A new feature, Deep Research, uses advanced reasoning to act as a research assistant. This feature explores complex topics and compiles reports, available in Gemini Advanced today. Google is also showcasing Project Astra and Project Mariner, built with Gemini 2.0. These prototypes highlight future AI capabilities in real-world applications.
Project Astra, a universal AI assistant, now converses in multiple languages and understands accents and uncommon words better. It uses Google Search, Lens, and Maps for everyday assistance. Improved memory and latency make it more personalized and efficient.
Project Mariner, an early research prototype, can understand and reason across web content in browsers. Evaluated against the WebVoyager benchmark, it achieved a state-of-the-art result of 83.5% as a single agent setup. Trusted testers are currently evaluating this prototype.
Google plans to integrate Gemini 2.0 across its products. The AI Overviews feature in Search, reaching 1 billion users, will be enhanced with Gemini 2.0’s advanced reasoning to handle complex queries. Early next year, Gemini 2.0 will expand to more Google products.
Built on custom hardware like the Trillium TPUs, Gemini 2.0 benefits from decade-long investments in AI innovation. These TPUs powered 100% of Gemini 2.0’s training and inference, now generally available to customers.
“Gemini 2.0 is about making information truly useful,” Pichai emphasized. This new era of AI promises transformative experiences, advancing science, technology, and everyday applications.
