Google Launches Gemini 2.0 Flash: Twice as Fast, Multimodal AI with Agentic Focus

Dec 11, 2024, 03:51 PM

Google has unveiled Gemini 2.0, its most advanced AI model to date, designed for the agentic era. The new model, Gemini 2.0 Flash, offers significant improvements over its predecessors, outperforming the Gemini 1.5 Pro on key benchmarks at twice the speed. Gemini 2.0 introduces enhanced multimodal capabilities, including text, image, and audio generation, as well as native tool use, 2D/3D spatial understanding, a 128k tokens context window, and a Multimodal Live API with audio and video streaming inputs. It offers multilingual native audio output and native image generation. The model enables developers to build AI agents capable of interacting with web browsers, controlling Chrome, moving the cursor, clicking buttons, and filling out forms. Gemini 2.0 is made available in an experimental phase to developers via the Gemini API, Google AI Studio, and Vertex AI. Google also unveiled Project Mariner and Project Astra, prototypes demonstrating the agentic capabilities of Gemini 2.0 in creating universal AI assistants and coding agents, including Jules, an AI code agent. Leading figures at Google, including CEO Sundar Pichai and Google DeepMind CEO Demis Hassabis, announced the release and highlighted the model's advancements in multimodality, performance, speed, and agentic focus.