Google has unveiled a major upgrade to its Gemini AI assistant, making its agentic Task mode available free, in an update that includes a faster AI model and new capabilities for making purchases and automating daily tasks.

Previously limited to paid subscribers, Gemini Task is rolling out free of charge in the U.S., Canada, Australia, India, Japan and South Korea. Google is also introducing advanced features in the U.S., including automated business calls, online payments and access to password-protected websites.

The release is a big competitive push in the agentic AI market, where rivals OpenAI and Anthropic charge for comparable capabilities. By eliminating subscription fees, Google is positioning Gemini to reach a far larger audience.

Central to the upgrade is Gemini 4 Hydrogen, Google’s recently introduced efficiency-focused LLM. While less powerful than the company’s flagship Gemini 4 Carbon, Hydrogen is built to deliver strong performance at lower computing costs. Google says the new model completed some data-intensive tasks up to 2.8 times faster than Gemini 3.8 Flash in early testing. It also supports parallel processing, allowing the assistant to perform multiple steps simultaneously rather than completing each action individually.

Among other tasks, Gemini can contact local businesses, check appointment availability and schedule a visit based on information in Google Calendar. The assistant uses Google Duplex to make phone calls and identifies itself as an AI when speaking with businesses.

Josh Woodward, Google’s VP responsible for Gemini, described using the assistant to schedule maintenance for his family’s 2019 Honda Odyssey. Gemini contacted two dealerships simultaneously and completed the scheduling process in less than five minutes.

E-Commerce Features

The upgrade also expands Gemini’s e-commerce capabilities. Using Google Pay and the Agent Payments Protocol (AP2), Gemini can compare products, identify discounts and help complete purchases. Users can establish spending limits and must approve transactions or provide authorization in advance. Google is working with Mastercard and Visa to support tokenized payments, with American Express support planned. The technology substitutes digital tokens for actual card numbers, reducing the risk of exposing customers’ financial information during transactions.

The company also plans to integrate the Universal Commerce Protocol (UCP), a standard designed to improve communication between AI agents and online retailers.

Gemini’s expanded access to personal and financial information also raises major security concerns. An AI agent that can access financial accounts, passwords and personal correspondence requires robust safeguards.

Google says each Gemini Task operates within a dedicated, hardware-isolated virtual machine using Google Cloud’s Agent Compute Engine. Additional protections include prompt-injection filtering, Google Safe Browsing and user controls governing application access.

Another enhancement is Gemini’s access to personal information stored across Google’s services. Through integrations with Gmail, Google Password Manager and other applications, the assistant can examine bills, organize appointments and access websites with user permission. It can also receive instructions through RCS messages or email.

Google says Gemini currently connects with more than 60 applications, expanding the range of services users can manage through the assistant.

Still, the company acknowledges that some tasks may fail or require users to complete individual steps manually, particularly when websites restrict automated activity. The most important test will be whether Gemini can perform complex tasks reliably while protecting the sensitive information it needs to operate.

In any case, making these capabilities free will likely boost agentic AI adoption while putting pressure on competitors to reconsider their pricing.