Gemini 2.0 Flash (experimental) is a multimodal generative AI model developed by Google, introduced as the initial offering in the Gemini 2.0 model family. Designed for the "agentic era," it emphasizes high-speed performance and low-latency interaction while maintaining high-quality reasoning and multimodal understanding. It was released as an experimental preview to allow for the development and testing of real-time applications.
The model supports a 1-million-token context window and natively processes diverse inputs including text, images, video, and audio. Key updates in this iteration include native tool-use capabilities, allowing the model to interact directly with tools such as Google Search and code execution. It also introduces support for multimodal outputs, including natively generated images and steerable text-to-speech functionality.
Architecturally, Gemini 2.0 Flash is engineered to deliver a significant performance boost over its predecessor, Gemini 1.5 Flash. According to technical reports, the model achieves quality levels comparable to or exceeding the larger Gemini 1.5 Pro while operating at approximately twice the speed. It is optimized for high-throughput, cost-efficient workflows required for sophisticated AI agents and large-scale deployments.