Google DeepMind has unveiled a new demonstration of Gemini 3.1 Flash-Lite, highlighting the model’s ability to generate complete websites in real time. The “Flash-Lite Browser” experiment, available through Google AI Studio, creates interactive and functional webpages from scratch as users type prompts and navigate through the interface.
In a video shared by Google DeepMind, developers showcased the model producing around 2,000 tokens of code in roughly four seconds, then regenerating full webpage layouts as users interacted with different elements. The demonstration included examples such as a mock social media interface and an interactive plant care guide, emphasizing the system’s capacity to produce diverse and responsive content without relying on pre-existing templates or data.
Gemini 3.1 Flash-Lite, which entered preview earlier this month through Google AI Studio and Vertex AI, is designed for speed and affordability. The model offers a 2.5 times faster time to first output token and 45 percent higher overall generation speed compared with its predecessor, Gemini 2.5 Flash. It is priced at $0.25 per million input tokens and $1.50 per million output tokens, with support for a one-million-token context window and adjustable “thinking levels” for performance optimization.
Developers testing the model noted its strong capabilities in generating front-end code with minimal adjustments, particularly for rapid prototyping. However, some reported reliability issues during prolonged interactions, such as frozen responses. The experiment also raises broader questions about how dynamically generated web experiences would handle persistence, analytics, and consistency in real-world applications.
While the “Flash-Lite Browser” remains experimental, the demonstration underscores Google’s continuing push to make AI generation faster, more efficient, and more accessible for developers building interactive digital experiences.
Read Article: India to Replace Form 16 With New Form 130 From April 2026
