On May 19 (local time), Google strengthened its offensive in the AI market at its annual developer conference, Google I/O, by unveiling a next-generation lightweight AI model, "Gemini 3.5 Flash," an active AI agent, "Gemini Spark," and a next-generation multimodal model, "Gemini Omni."


"Gemini 3.5 Flash" demonstrated performance surpassing the top public models such as Gemini, GPT, and Claude on model context protocol (MCP)—an agent framework—and financial analysis benchmarks. As a lightweight model, it is up to half to a third less expensive than other advanced models, and its output speed is up to four times faster, making it highly suitable for enterprise environments. Sundar Pichai, CEO, explained that companies are facing budget constraints due to increased AI usage, adding, "If a company using 1 trillion tokens per day switches 80% of its workload to the Flash model or similar, it can save over $1 billion annually."

Sundar Pichai, CEO of Google, is giving a speech at the Google I/O event on the 19th (local time). Photo by Getty Images Yonhap News

Sundar Pichai, CEO of Google, is giving a speech at the Google I/O event on the 19th (local time). Photo by Getty Images Yonhap News

View original image

The company also introduced the active agent "Gemini Spark." Even when users are not actively using their devices, the agent can handle tasks such as email summarization, briefing creation, and repetitive work 24 hours a day. Google's iconic search service will also undergo a major overhaul. In addition to text, users will be able to search using images, videos, and files, and AI-powered visual content and widgets will be generated in the search results. The "AI Overview" feature will connect to a chatbot-like "AI Mode," providing an interactive search experience. A "Universal Cart" feature that handles everything from price tracking to payment is also being introduced.


The company also unveiled "Gemini Omni," capable of processing all types of input and output, including text, images, audio, and video. This service aims to fill the gap left by OpenAI's recent discontinuation of its video-generating AI, "Sora." In the security field, Google will compete in earnest with Anthropic's "Claude Mythos."


CEO Pichai announced the upcoming release of the "CodeMender" API, a security agent, emphasizing, "We internally confirmed that even using the older Gemini 3.1 Pro model, it is possible to detect 80% to 90% of vulnerabilities in the cybersecurity sector."


To expand its market share, Google has also played the price reduction card. The monthly fee for the Ultra plan, previously at $249.99, has been lowered to $100–$200, matching the top-tier plans of OpenAI and Anthropic.



Meanwhile, on this day, shares of Google's parent company Alphabet closed down 2.23% in regular trading, due to factors such as rising bond yields and profit-taking.


This content was produced with the assistance of AI translation services.

© The Asia Business Daily(www.asiae.co.kr). All rights reserved.

Today’s Briefing