Google unveils Gemini Spark agent, 3.5 Flash, and Omni video model at I/O
Gemini app now has 900M monthly users; Spark agent works even when your laptop is closed.
At Google I/O, CEO Sundar Pichai unveiled a wave of AI products, highlighting the company’s transition into an "agentic" era. The centerpiece is Gemini Spark, a cloud-based AI agent powered by Gemini 3.5 that performs background tasks like organizing meeting notes, chats, and emails—even with the laptop closed. It asks permission for high-stakes actions like sending emails or purchases. Limited beta starts today for select testers, rolling out to Google AI Ultra subscribers in the U.S. later.
Alongside Spark, Google released Gemini 3.5 Flash, its strongest coding and agentic model, now default in the Gemini app and AI mode on Search. It claims 4x faster inference than competitors. Also new is Gemini Omni, a multimodal model for creating realistic videos from text, images, or audio—understanding gravity, kinetic energy, and fluid dynamics. Omni Flash launches for subscribers today and free on YouTube Shorts. Google also announced that OpenAI, Kakao, and Eleven Labs will adopt its SynthID watermarking. The Gemini app now boasts 900M monthly active users, more than double last year.
- Gemini Spark: cloud-based agent that works 24/7, even when devices are off, and asks permission for sensitive tasks.
- Gemini 3.5 Flash: 4x faster than competitors, strongest agentic and coding model, now default in Gemini app and Google Search AI mode.
- Gemini Omni: generates physics-aware video from any input (text/image/audio) with SynthID watermark; free on YouTube Shorts this week.
Why It Matters
Google pushes proactive AI agents and faster models, accelerating practical automation for knowledge workers and creators.