Google on Tuesday introduced Gemini 3.5 and opened the series with 3.5 Flash, a model it says is built to combine frontier intelligence with action. The company said 3.5 Flash is available now to billions of people globally, while 3.5 Pro is already being used internally and will roll out next month.
The launch gives Google a fresh push around gemini spark, its pitch that the next wave of AI should not only answer questions but carry out work. 3.5 Flash is the first public piece of that plan, and Google is framing it as the fastest way into a family of models focused on agentic workflows and coding.
The numbers are the point. Google said 3.5 Flash beats Gemini 3.1 Pro on Terminal-Bench 2.1 at 76.2 percent, GDPval-AA at 1656 Elo and MCP Atlas at 83.6 percent. It also leads in multimodal understanding with 84.2 percent on CharXiv Reasoning. The company said the model runs four times faster than other frontier systems when measured by output tokens per second and often costs less than half as much.
That matters because Google is trying to sell speed and price as more than technical bragging rights. The company said 3.5 Flash can finish long-horizon agentic tasks in a fraction of the time, from multi-step workflows to coding jobs that would normally drag on. In one example, it said the model automatically renamed and categorized unstructured assets based on dynamic criteria. In another, it said two agents used the model to synthesize the AlphaZero paper and produce a fully playable game in six hours.
Google also said 3.5 Flash transformed a messy legacy codebase to Next.js using the Antigravity harness, a sign that the company is pushing hard on developer tools as well as consumer-facing AI. Under supervision, the model can execute multi-step workflows and coding tasks while sustaining frontier performance, and Antigravity allows it to deploy collaborative subagents for more demanding use cases.
The context is broader than one model release. Gemini 3.5 is being presented as a new family centered on agentic work, with 3.5 Flash built on the multimodal foundation of Gemini 3. The company is positioning the model as a way to deliver frontier-level intelligence without the lag and cost that can keep users and developers from handing over real work.
Google said the system can also generate richer, more interactive web UIs and graphics. On AI Studio, it can create interactive animations for a research paper, turn a plain-text description into interactive hardware, execute multiple concepts in parallel for a school fundraiser branding project and generate different UX approaches for a checkout flow in 60 seconds.
The tension in the launch is obvious. Google is promising a model that is faster, cheaper and more capable, but its own examples show how much supervision and orchestration still sit behind the curtain. The company is not yet putting 3.5 Pro in front of the public, and the real test will be whether the Flash version can carry the load once more users start pressing it into day-to-day work.
For now, Google has put its first marker down. 3.5 Flash is live, 3.5 Pro is close behind, and Gemini Spark is no longer just a pitch about what AI might do next.
