- Google unveiled Gemini 3.5 Flash, a faster AI model outperforming its predecessor and rivals
- The company introduced Gemini Spark, an AI assistant for multitasking across Google apps
- Google launched Docs Live, enabling voice-based real-time document creation and editing
Google unveiled new artificial intelligence models, software tools and custom AI chips at its annual developer conference on Tuesday, as the company expanded its push into AI-powered consumer and enterprise products.
Chief Executive Officer Sundar Pichai introduced Gemini 3.5 Flash, a new AI assistant called Gemini Spark, Docs Live voice tools and the TPU 8t and 8i chips during the Google I/O 2026 event. The company also launched a Google AI Ultra subscription plan priced at $100 a month, while reducing the price of its top-tier Ultra plan from $250 to $200 a month.
The announcements underline Google's effort to integrate AI tools across its products and services, while competing with rivals including OpenAI in generative AI, software automation and AI infrastructure.
Gemini 3.5 Flash
Pichai described Gemini 3.5 Flash as "the first in a series of models combining frontier intelligence with action".
He said the model performs better than its predecessor, Gemini 3.1 Pro, across most benchmarks. According to Google, Gemini 3.5 Flash generates 289 output tokens per second, compared with 71 tokens per second for OpenAI's GPT 5.5.
Google also introduced "Google Antigravity", an AI agent system designed to automate software development. The company said its internal teams are already using the tool to speed up coding and development processes.
Gemini Spark
Google introduced Gemini Spark, an AI assistant designed to complete tasks across apps including Gmail and Google Drive.
Pichai said the assistant is meant to "help you navigate digital life, takes action on your behalf, and under your direction".
The company demonstrated Spark planning a block party by compiling RSVPs, sending emails, creating reminders and preparing a Google Slides presentation. Google said the assistant can continue running tasks in the background even after users close their laptops.
Spark can also generate emails using a "ghostwriter" feature that reflects a user's writing style.
Google said users will be able to access Spark through the Gemini app, with email and chat integration expected later. The feature will be available in beta for Google AI Ultra subscribers in the US from next week. It will support both Android and iPhone devices.
ALSO READ: Sales Jobs To Survive? How AI May Redraw India's Career Map
Omni And Content Credentials
Demis Hassabis showcased Google Omni, a media-generation tool capable of producing audio, video, images and text.
Google said Omni can edit videos using conversational commands and allows users to adjust styles, scenes and elements in existing videos. The feature will be available through Gemini, Google Flow and YouTube Shorts.
The company also announced expanded content credential verification tools aimed at identifying whether media was generated or edited using AI.
Hassabis said users can verify media by using Circle to Search. He added that companies including Nvidia and OpenAI have adopted the standard, while Google plans to expand the system to additional partners.
Docs Live
Google also introduced Docs Live, a voice-based document creation tool integrated with Gemini.
The feature allows users to speak naturally while Gemini converts spoken instructions into a Google Doc in real time.
"In the future you will be able to create new docs and edit them directly, all with your voice," Pichai said.
Google demonstrated a user verbally dictating a speech while making changes and additions in real time. Docs Live was also shown retrieving information from Google Drive during the session.
TPU 8t And 8i
Google unveiled two new AI chips, TPU 8t and TPU 8i, built for training and inference workloads.
The company said TPU 8t delivers three times the computing power of the previous generation and allows AI training workloads to scale beyond the limits of a single data centre.
Google also said the chips can deliver inference speeds of up to 1,500 tokens per second and provide twice the energy efficiency and performance per watt compared with earlier versions.
ALSO READ: Microsoft Plans To Operationalise Its Largest Data Centre In India By Mid-2026
Essential Business Intelligence, Continuous LIVE TV, Sharp Market Insights, Practical Personal Finance Advice and Latest Stories — On NDTV Profit.
