The new version of Gemini makes the model more suitable for planning, remembering, and acting under user guidance.
Effective immediately, an experimental version of Gemini 2.0 Flash is available. Google describes this version of Gemini as a workhorse with low latency and improved performance. Developers can use Flash directly through the Gemini API in Google AI Studio and Vertex AI. In addition, a chat-optimized experimental version of Gemini 2.0 Flash is available for users and is easily activated via the drop-down menu.
Gemini 2.0 features enhanced reasoning capabilities. These allow the model to better handle complex queries consisting of multiple parts, multimodal queries, and programming challenges.
Agent projects
Google is also taking the next step in responsible AI development. The company is launching several research prototypes and experiments to explore how agentic experiences can operate safely and under practical conditions. These include Project Astra, which explores universal AI agents with multimodal reasoning capabilities.
Another research prototype using Gemini 2.0 is Project Mariner. This project is based on the hypothesis that agents are valuable in performing complex tasks. It explores how agents and humans can work better together in the future.
On the agent front, Google also introduces an AI-powered programming agent called Jules. These agents integrate seamlessly into the GitHub workflow and are designed to work optimally with the tools developers use daily.