Google I/O 2026: Gemini 3.5 Flash and Gemini Spark Debut as Gemini App Hits 900 Million Users

Google used its annual I/O developer conference on May 19, 2026 to announce Gemini 3.5 Flash — a frontier-class model designed for fast, low-cost agentic inference — and Gemini Spark, a 24/7 personal AI agent. The Gemini app has surpassed 900 million monthly active users, more than doubling from 400 million at last year's I/O, while Google Search's AI Mode exceeded 1 billion monthly active users.

What's new

Gemini 3.5 Flash: Google introduced this as "our first in a series of models combining frontier intelligence with action." The company positioned it as a frontier-level model — comparable in intelligence to the best available models — but running at substantially faster throughput and lower cost. It is designed specifically for agentic workloads where cost and latency matter alongside capability.

Gemini Spark: A personal AI agent running 24/7 on dedicated virtual machines within the Gemini app. CEO Sundar Pichai described it at the keynote: "I'm particularly excited for Gemini Spark, your personal AI agent in Gemini app that helps you navigate your digital life, taking action on your behalf and under your direction."

Gemini Omni Flash: A new model capable of generating output in any modality from any input type, including video generation.

Scale milestones:

• Gemini app: 900 million monthly active users, up from 400 million one year ago
• Google Search AI Mode: exceeded 1 billion monthly active users

Infrastructure investment: Google announced capex is increasing to approximately $180–190 billion annually, up from $31 billion in 2022, with new TPU 8t and 8i custom chips.

Gemini app product updates:

• Proactive background assistance
• Agentic search with generative UI and custom dashboards
• New features: paper note digitization, in-app file generation

Context

This year's I/O arrived after what Pichai called "an extraordinary year since our last I/O, a period of relentless shipping, technology advances and hyper progress." The keynote's central theme — "the agentic Gemini era" — signals a deliberate shift in product narrative from conversational AI to action-taking AI.

The user numbers establish Google's scale advantage. Reaching 900 million Gemini app monthly users in roughly two years from general availability is a consumer adoption rate that rivals Google's core products. AI Mode in Search at 1 billion monthly users shows that AI interfaces have crossed from early adopter into mainstream consumer behavior at Google's distribution scale.

Gemini Spark's architecture — dedicated VMs per user, running continuously — is a significant infrastructure commitment. It implies ongoing compute cost at per-user granularity, a bet that persistent agentic AI will become a core product rather than an optional add-on.

Why it matters

Gemini 3.5 Flash is the most consequential announcement for developers. If the speed and pricing claims hold under third-party benchmarks, it would substantially change the economics of agentic applications — particularly those with high per-interaction compute costs. A fast, frontier-quality model at lower price changes what's viable to build.

Gemini Spark represents Google's concrete bet on always-on AI agents. Multiple companies are exploring this architecture, but Google is deploying it at consumer scale now, with 900 million potential users as the addressable base.

Google's data flywheel is worth noting. Every interaction across 900 million Gemini app users and 1 billion AI Mode search users generates observational data about real-world AI use — a structural advantage in model improvement and product iteration that API-first competitors cannot replicate at equivalent scale.