Google unveiled a comprehensive overhaul of its artificial intelligence offerings at its annual I/O developer conference on Tuesday, introducing new AI models and fundamentally redesigning its search interface for the first time in 25 years. The company announced Gemini 3.5 Flash, a new AI model that Google claims outperforms its previous flagship while running four times faster and at significantly lower cost.

The tech giant redesigned its iconic search box to accept multimodal inputs including text, images, videos, and files, moving away from the traditional keyword-based approach. The new interface merges AI Overviews with AI Mode into a seamless conversational experience, representing what Google called the biggest upgrade to search since its debut over 25 years ago. AI Mode has reached one billion monthly users in its first year, with query volume doubling each quarter.

Google also introduced Gemini Spark, an always-on AI agent that runs persistently in Google Cloud to handle tasks like drafting emails and monitoring inboxes even when users' devices are offline. The company announced the Agent Payments Protocol to enable secure autonomous purchases within user-defined limits. Additional announcements included Gemini Omni, a multimodal model capable of generating video from various inputs, and expanded integration with third-party services.

Separately, OpenAI co-founder Andrej Karpathy announced he is joining rival AI company Anthropic to focus on pretraining research. The 39-year-old researcher, who previously led Tesla's AI division and returned to OpenAI from 2023 to 2024, will build a team focused on using Anthropic's Claude model to accelerate pretraining research. Karpathy was one of OpenAI's original 11 co-founders and has been a prominent AI educator since leaving the company.

Google's announcements come amid intense competition in the AI space, with the company expecting capital expenditures of approximately $180-190 billion in 2026 to support its AI infrastructure. The company processes over 3.2 quadrillion tokens monthly across its services, a seven-fold increase from the previous year. Google claims enterprises could save over $1 billion annually by shifting workloads to its more efficient models.