Google Recasts Gemini as an AI Agent Platform With Flash and Omni
Google is pushing to turn Gemini from a chatbot into the operating layer for its AI ambitions, unveiling a wave of new models and features aimed squarely at rivals ChatGPT and Claude.
The centerpiece is Gemini 3.5 Flash, the first in Google's latest series pairing frontier-level intelligence with the speed the Flash line is known for. Google says it rivals larger flagship models while outperforming its own previous generation on demanding coding and agentic benchmarks — the kind of multi-step tasks where an AI system carries out work rather than just answering questions.
Alongside it, Google introduced Gemini Omni, a model built to generate output from any input, starting with video. The company says Omni pairs Gemini's reasoning with a sharper grasp of physics — gravity, kinetic energy and fluid dynamics — an area where AI video generators have often stumbled.
The Gemini app itself has been rebuilt from the ground up with a new design language, adding a 'Daily Brief', a personal AI agent called Gemini Spark and a redesigned interface heavy on fluid animation and haptic feedback. Google is also courting power users with a new premium tier offering far higher usage limits for developers, technical leads and advanced creators.
The strategy is clear: make Gemini the default AI layer woven through Google's products and, increasingly, an agent that acts on a user's behalf. In a market where OpenAI and Anthropic are moving fast, Google is betting that tighter integration and multimodal reach can win over users who now have more capable assistants to choose from than ever.
