Google launches Gemini Spark agent

Google used I/O 2026 to present a different model of assistant: one that keeps running after the user closes the app. Gemini Spark, unveiled on May 19 alongside Gemini 3.5 Flash, is described by Google and outside reports as a persistent agent that can take on tasks in the background rather than wait for a prompt every time. That matters because most consumer AI tools still behave like sessions. Spark is being positioned as a process. ### What exactly did Google announce? Google said on May 19 that Gemini 3.5 is a new family of models built to combine “frontier intelligence with action,” with Gemini 3.5 Flash as the fast model aimed at agentic workflows. In a separate Gemini app update, Google said the app is becoming more “agentic” and pitched 3.5 Flash as the engine for faster task execution. Outside coverage of the I/O event described Gemini Spark as the clearest product expression of that push. TechCrunch reported Spark as a new personal assistant built from Gemini models and Google’s Antigravity agent framework, while The Next Web said it runs 24/7 on cloud virtual machines and can take requests through Gmail. (blog.google) ### Why is Spark different from a normal chatbot? Gemini Spark is being described as always on. DataCamp said the system keeps running on Google’s cloud infrastructure even after a laptop is closed or a phone is locked, and SiliconANGLE described it as a 24/7 assistant for daily digital tasks. That changes the operating model. Uddhav Bhople, writing on Dev.to after I/O, said the shift is from “agents as subroutines you call” to “agents as processes that run continuously.” That is commentary, not a product spec, but it captures the distinction Google is trying to establish around persistence and asynchronous work. (techcrunch.com) (datacamp.com) ### What jobs is Google aiming Spark at? The early descriptions center on routine digital work. The Next Web said Spark can browse through Chrome, accept tasks through email and keep working in the cloud for AI Ultra subscribers. DataCamp said Google framed it around monitoring Gmail, managing Calendar and drafting documents, with commerce actions planned later. (dev.to) Google’s own language around Gemini 3.5 Flash supports that direction. The company said the model is designed for “real-world agentic workflows,” and Google’s I/O developer materials tied the launch to a broader move “from prompts to action.” ### Why does this matter beyond Google’s product lineup? Google’s launch adds a mainstream example of a persistent assistant. (thenextweb.com) If the model is no longer “ask, wait, answer” but “assign, monitor, continue,” then product teams will need clearer ways to show what an agent is doing, what it has permission to do and when a human must approve the next step. That is an inference from the product design, based on the way Spark is described as continuing work over time. (blog.google) Google also tied the launch to a wider agent platform story. Its I/O roundup said Antigravity had moved beyond tools that help people write toward agents that help them act, placing Spark inside a larger push across Gemini, Search and developer tooling. ### When can people use it? MakeUseOf reported that Gemini Spark is rolling out to trusted testers first, with a beta for U.S. (thenextweb.com) Google AI Ultra subscribers planned next. The Next Web likewise tied access to AI Ultra. Google’s public I/O posts highlighted the Gemini app updates and Gemini 3.5 launch on May 19, but did not, in the material reviewed here, publish a broader public release date for Spark. (makeuseof.com) (blog.google)