Google has unveiled Gemini 2.0, its latest artificial intelligence model designed to understand and act on information across multiple formats. CEO Sundar Pichai and the Google DeepMind team announced the release as part of what they describe as an "agentic era" of AI development.
Gemini 2.0 Flash, the first model in the 2.0 family, is being released as an experimental version and can be accessed by Gemini users by selecting it in the model drop-down menu on desktop and mobile web platforms, with mobile app availability coming soon. The model builds on its predecessor with enhanced performance and new capabilities, including the ability to generate images and multilingual audio alongside text responses.
For developers, Gemini 2.0 Flash is available through the Gemini API in Google AI Studio and Vertex AI. The company says the model outperforms the previous 1.5 Pro version at twice the speed, with general availability planned for January along with additional model sizes.
Google is also showcasing several research prototypes. Project Mariner, an experimental Chrome extension, explores browser-based task completion. Project Astra, which has been in testing on Android phones, now features improved dialogue capabilities, including multilingual conversations and integration with Google Search, Lens, and Maps. The company also introduced Jules, an experimental AI coding agent that works within GitHub workflows.
A new feature called Deep Research, which leverages advanced reasoning and long context capabilities to explore complex topics and compile reports, is now available in Gemini Advanced.
Google emphasised its gradual approach to development, noting that these technologies are being tested with trusted testers and external experts, with extensive risk assessments being conducted. The company plans to expand Gemini 2.0 to more Google products early next year.
As we rapidly approach 2025 and the Year of Agentic AI Gemini 2.0 offers a glimpse of what is to come.