Benchmark test results published by reputable AI research organizations or tech media outlets

Google's Gemini 2.0: Advancing AI with Multimodal Capabilities and Agentic Features

Dec 12, 2024, 03:26 PM

Google has launched Gemini 2.0, its latest AI model, introducing significant advancements in multimodality and performance. The model can now process and generate images, audio, and text, enhancing its capabilities to understand and interact with various forms of media. Gemini 2.0 Flash, a version optimized for speed and efficiency, outperforms its predecessor, Gemini 1.5 Pro, in key benchmarks while offering twice the speed. This new model supports multimodal outputs, including native image and audio generation, and can access external tools like Google Search, Maps, and Lens. Google aims to integrate Gemini 2.0 into its ecosystem, including search, YouTube, and Android, to usher in an 'agentic era' where AI can perform complex tasks independently. The launch includes new features like Deep Research, an AI-driven research assistant, and several agentic AI projects like Project Astra, which can analyze real-time video streams and interpret data from Google Maps; Project Mariner, a prototype for automating browser-based tasks; and Jules, an AI coding agent for developers. Additionally, Gemini 2.0 supports advanced reasoning, long context understanding, complex instruction following, compositional function-calling, native tool use, and improved latency.