Google launches Gemini 3.5 Flash

Google used its I/O conference on May 19 to introduce Gemini 3.5 Flash, a new model the company is pitching as a faster, cheaper option for coding and AI agent workloads. Google said the model is designed for “sustained frontier performance” on agentic and coding tasks, placing it between raw capability and production economics in a market where developers are weighing speed, latency and cost as much as benchmark scores. The launch was part of a broader I/O push aimed at developers, alongside updates to the Gemini API, AI Studio and Google’s agent tooling. Google’s release notes show Gemini 3.5 Flash is already generally available through the Gemini API. ### What exactly did Google launch at I/O? Google said Gemini 3.5 Flash is the first model in its new 3.5 family and described it as a system that combines “frontier intelligence with action.” In the company’s I/O roundup, Google paired the launch with Gemini Omni and updates to Antigravity, its agent-focused development platform. The Gemini API documentation lists Gemini 3.5 Flash as a stable model and calls it Google’s “most intelligent model for sustained frontier performance on agentic and coding tasks.” The release notes say the generally available version, named `gemini-3.5-flash`, was released on May 19. (blog.google) ### Why is Google stressing speed and cost this time? Google’s developer highlights post said Gemini 3.5 Flash “outperforms Gemini 3.1 Pro across almost all benchmarks” while running “four times faster than other frontier models.” The company framed that trade-off as central to real-world agentic workflows, where models may need to make repeated tool calls, maintain state and respond with low latency. (blog.google) (ai.google.dev) The Gemini API model pages also place 3.5 Flash alongside other lower-cost and high-volume options, including Gemini 3 Flash and Gemini 3.1 Flash-Lite. That product lineup suggests Google is segmenting its models more explicitly by workload and operating cost, an inference supported by how the company labels 3.5 Flash as a higher-intelligence Flash-tier model rather than a full Pro replacement. (blog.google) ### How is Google positioning it against Gemini 3.1 Pro? Google’s own materials compare Gemini 3.5 Flash directly with Gemini 3.1 Pro, a model the company describes as its advanced reasoning and problem-solving system. In the API documentation, Gemini 3.1 Pro is presented as the company’s top model for complex reasoning, while 3.5 Flash is presented as the better fit for sustained coding and agentic execution. (ai.google.dev) That positioning matters because Google is not saying 3.5 Flash replaces Pro on every dimension. Instead, the company is presenting a workload-specific claim: developers building coding assistants, tool-using agents and interactive systems may get better throughput and lower operating costs without moving to a larger, slower model. ### Where does this fit in Google’s broader I/O pitch to developers? (ai.google.dev) Google’s I/O 2026 developer post tied Gemini 3.5 Flash to a wider shift “from prompts to action.” The company announced Managed Agents in public preview at the same time, describing them as autonomous, stateful agents that run in secure Google-hosted Linux sandboxes. (blog.google) Sundar Pichai’s keynote post and Google’s Gemini app update both linked 3.5 Flash to more agentic consumer and developer experiences. Google said the Gemini app now reaches more than 900 million monthly users across 230 countries and more than 70 languages, giving the company a large distribution base as it rolls new models into products and developer tools. ### What should developers watch next? (blog.google) May 19 is the key date in Google’s release notes because it marks both the general availability of Gemini 3.5 Flash and the public preview of Managed Agents. Developers can track model status, deprecations and rollout changes through the Gemini API model pages and changelog. Google said in its main Gemini 3.5 post that it had already been using 3.5 Flash internally and was looking forward to broader rollout “next month.” The next concrete milestone is how quickly the model appears across Google AI Studio, Vertex AI and higher-level products tied to the Gemini and Antigravity stack. (blog.google 1) (blog.google 2) (ai.google.dev)