Google I/O 2026: Gemini 3.5 Flash and the Spark agent

Google answers OpenAI and Anthropic

At its I/O 2026 conference (May 19), Google expanded the Gemini family with two announcements that matter: a new lightweight model, Gemini 3.5 Flash, and a personal agent, Gemini Spark. The stated goal is clear — keep pace with OpenAI and Anthropic on the AI-agent front.

Gemini 3.5 Flash — fast and built for agents

Flash is positioned as a 'frontier'-performance model for agents and coding, able to handle long, multi-step tasks. Its pitch: it generates tokens roughly 4x faster than other models in its tier. On the numbers, it beats Gemini 3.1 Pro on demanding benchmarks like Terminal-Bench 2.1 (76.2%), MCP Atlas (83.6%) and CharXiv Reasoning (84.2% multimodal).

Pricing is $1.50 / $9.00 per million tokens (input / output) — about 3x its predecessor, but Google frames it as cheaper than Claude Opus 4.7 or GPT-5.5 on most workloads. Flash is available now in the Gemini app, AI Mode in Google Search, and for developers via AI Studio and Antigravity.

Gemini Spark — the agent that runs even with your device off

Spark is an 'always-on' personal agent that runs in the cloud: it keeps working on your tasks even when your phone or laptop is off. Built on the Antigravity platform, it plugs into Google Workspace (Gmail, Docs, Calendar) and third-party tools via the MCP protocol. Rollout starts in beta for Google AI Ultra subscribers in the U.S., with a summer roadmap: text or email Spark, create sub-agents, and authorize payments by setting a budget and merchants.

Gemini Omni — the multimodal that generates video

Third big announcement: Gemini Omni, Google's native multimodal model, which handles text, image, video, audio and code in a single architecture where every modality is treated equally. Its starting point: generating video from any input, combining Gemini's intelligence with Google's best generative models. The Omni Flash version is already rolling out to Google AI Plus, Pro and Ultra subscribers worldwide, via the Gemini app and Google Flow, and free in YouTube Shorts Remix and the YouTube Create app (18+).

Android XR glasses and Search switching to Flash

Google also confirmed its Android XR smart glasses: the first audio glasses arrive this fall, made with Samsung and Qualcomm, design by Gentle Monster and Warby Parker. They give always-on access to Gemini, with responses privately spoken into your ear, and work even paired with an iPhone (photos, music, calls, apps).

On search, Google's 'AI Mode' has passed one billion monthly users and is switching to Gemini 3.5 Flash as the default model globally. In other words, the new model isn't confined to developers: it becomes the AI-answer engine that hundreds of millions of people use without even knowing it.

Our take

The 2026 trend holds: the big three (Google, OpenAI, Anthropic) are all pushing toward agents that can chain tasks autonomously. Spark goes further than most by running server-side around the clock — handy, but it also raises trust questions once you let it authorize payments. For personal or creative use, Flash is the most directly useful announcement: fast, multimodal, and available right away.