Google battles OpenAI and Anthropic with new Gemini models and AI agents

The significant announcements took place at the annual Google I/O developer conference on Tuesday. This event comes at a crucial time as the market focuses heavily on the rising valuations of competitors OpenAI and Anthropic.

GEMINI 3.5 FLASH DEBUTS WITH FASTER PERFORMANCE

The centerpiece of the tech giant’s current strategy revolves around Gemini 3.5 Flash, which joins the existing suite of tools. This lighter-weight option delivers cutting-edge capabilities while costing half or nearly one-third of the price of comparable frontier models.

CEO Sundar Pichai stated that this new addition performs remarkably fast during a news briefing with reporters. Google has already made Gemini 3.5 Flash the default model for its global search AI mode and the dedicated Gemini application.

Users do not have to trade quality for latency anymore according to an official blog post. Furthermore, enhanced cybersecurity defenses ensure that the system is less likely to generate harmful content or mistakenly refuse safe queries.

LATEST AGENTIC SERVICES AND FUTURE PRO MODELS

The heavier-weight version known as Gemini 3.5 Pro is currently being utilized for internal purposes. This particular model will not be ready for a wider distribution until next month.

Google also announced Gemini Spark, which operates as a new general-purpose AI agent within the Gemini application. This agent can reason across information in connected apps to take action on behalf of the user.

The company aims to help users navigate their digital lives through this virtual assistant under direct human guidance. Gemini Spark is starting in beta next week for trusted testers and Google AI Ultra subscribers.

INVESTMENT EXPECTATIONS AND WORLD MODEL OMNI

Wall Street wants Google to show deeper product integrations following the recent skyrocketing capital spending by the firm. Creating advanced AI agents could serve as a primary way to fulfill these financial expectations.

Expectations for artificial intelligence continue to grow after Anthropic recently released its powerful Mythos model. That rival model reportedly found thousands of previously unknown vulnerabilities within global software infrastructure.

Google responded by adding Omni to its portfolio, which serves as a world model designed to simulate physical environments. This technology predicts what happens next based entirely on the actions of a user.

OMNI CAPABILITIES IN VIDEO AND IMAGE EDITING

World models are heavily researched by DeepMind and typically find use cases within robotics and gaming. Omni will actively support both image and audio inputs across multiple platforms.

The new world model will function within Flash, the Gemini App, Google Flow, and YouTube shorts. Users can utilize Omni to edit videos and create more realistic imagery through simple prompts.

An official post explains that people can take a shot video and ask Omni to change what is happening. The artificial intelligence can edit the ongoing action or add entirely new characters and objects to the footage.