© Google
Google unveiled Gemini 3.5 Flash, a lightweight, fast, and cost-effective model, but one that would show similar, if not superior, performance compared to top models. As expected, it's about agentive AI. Gemini 3.5 Flash doesn't just answer questions; it acts on your behalf and executes actions.
During the conference Google I/O 2026, Google unveiled Gemini 3.5 Flash, a lightweight, fast, and cost-effective model, but one that would show similar, if not superior, performance compared to top models. As expected, it's about agentive AI. Gemini 3.5 Flash doesn't just answer questions; it acts on your behalf and executes actions.
A lightweight model as powerful as some top-tier models
At the Google I/O 2026 conference, Google introduced the new Gemini 3.5 model series, presenting Gemini 3.5 Flash, a new model that « combines top-tier intelligence with ultra-quick performance, operating up to four times faster than comparable models ». Gemini Flash aims to be a lighter and faster model, but it would actually be as powerful as some professional models capable of deep reasoning. Sundar Pichai, Google's CEO, explains that Gemini 3.5 Flash « shows better performance in almost all performance tests » compared to the 3.1 Pro version.
In addition to being fast and lightweight, Gemini 3.5 Flash would therefore demonstrate impressive performance, as Google explains: « It is our most powerful model in terms of agency and coding, surpassing Gemini 3.1 Pro on demanding benchmarks such as Terminal-Bench 2.1 (76.2 %), GDPval-AA (1656 Elo), and MCP Atlas (83.6 %), and standing out for its multimodal understanding (84.2 % on CharXiv Reasoning). In terms of tokens per second, it is four times faster than other top-tier models ».
An agentive AI
Gemini 3.5 Flash is designed to answer your questions, but also to act. During the presentation of this new model, Google particularly emphasized its agentive dimension: « it reliably executes complex workflows and coding tasks while maintaining optimal performance ». The American company clarified: « What previously took a developer days or an auditor weeks, 3.5 Flash can now accomplish in record time, often at less than half the cost of other top models. It plans, builds, and iterates quickly to solve concrete problems, whether it's developing new applications, maintaining codebases, or contributing to the preparation of financial documents ».
Google hopes that you will integrate Gemini 3.5 into your workflows, but also « into your daily life ». And when talking about everyday agentive AI, Google also introduced Gemini Spark, a personal AI agent, using Flash 3.5 technology, which: « operates 24/7 to accompany you in your digital life and act on your behalf according to your instructions ».
Gemini 3.5 Flash will be deployed throughout Google's ecosystem and is available now on the Gemini app, in developer APIs, and in Antigravity. Note that a model 3.5 Pro is planned, but it is currently in the testing phase.
👉🏻 Follow the latest tech news in real time: add 01net to your Google sources, and subscribe to our WhatsApp channel.
Source: Google













