Last Updated: 20 May 2026
Google has released Gemini 3.5 Flash, a new AI model built for coding and AI agents. It is now generally available in the Gemini API, and Google says it is made for long tasks, agent workflows, and fast coding loops.
The main idea is simple. Google wants Gemini to move beyond basic chatbot answers. With Gemini 3.5 Flash, the model is built to understand a task, plan the steps, use tools, write code, check progress, fix issues, and keep working until the job is ready for human review.
Google added Gemini 3.5 Flash to the Gemini API release notes on May 19, 2026. The company presents it as the generally available version of the model, with a clear focus on agent tasks and coding work. In simple terms, this means Google sees it as ready for real use, not just testing.
For developers, this makes the model easier to bring into actual products and workflows. Gemini 3.5 Flash can now be used through the Gemini API as part of Google’s public developer tools. It is not only an experiment shown on a stage. It is available for teams that want to build apps, coding assistants, and agent based systems.
Gemini 3.5 Flash also supports many types of input, which makes it more flexible. Google lists text, image, video, audio, and PDF as supported input types, while the output is text. This means developers can build tools that read different kinds of content and turn them into answers, summaries, instructions, or code.
That makes it useful for apps that need to read files, understand media, and produce written answers or code.
Gemini 3.5 Flash is built around agents, not only simple chat. An AI agent is a system that can receive a task, understand the goal, choose tools, create steps, and continue working with less human help. This makes the model more useful for jobs that need action, not just answers.
Google also launched Managed Agents in the Gemini API as a public preview. This gives developers a way to build and deploy autonomous agents with Google’s own infrastructure. These agents can work inside hosted Linux environments, which means developers can test more complex workflows without having to build every part of the system from zero.
The sandbox detail is especially important because it gives the agent a controlled place to work. It can run code, manage files, test changes, and handle tasks without directly touching the user’s real machine. That adds a safer layer between the AI agent and the developer’s local setup, which is useful when the model is allowed to take more actions.
What Gemini 3.5 Flash can do:
Antigravity is Google’s agent development platform and coding environment. It is designed for developers who want AI to do more than suggest code in a chat window. The Antigravity Agent in the Gemini API is powered by Gemini 3.5 Flash and uses the same core setup as the Antigravity IDE, which connects the model to a real coding workflow.
The Antigravity Agent can reason through a task, run code, manage files, and browse the web inside a Google hosted Linux sandbox. This means a developer can give it a job and let it work in a controlled environment. A single API call can start the agent, assign the task, and let the system begin working through the steps.
TNW said Google showed Gemini 3.5 Flash building an operating system through agents that worked on separate parts before joining their work together. The same coverage said Google released Antigravity 2.0 as a standalone desktop app with a CLI, SDK, and links to Google AI Studio, Firebase, and Android. This shows that Google is building Antigravity as a full workspace, not only as a model demo.
Google says Gemini 3.5 Flash is optimized for higher speed and lower cost while keeping strong performance for real tasks. The model page says it is useful for sub agent deployment, multi step workflows, and long tasks at scale.
The model supports code execution, function calling, structured outputs, file search, URL context, search grounding, and other tools inside the Gemini API. These features make it worthwhile because agents need tools to do work beyond simple chat.
Google presented Gemini 3.5 Flash as its strongest model yet for coding and autonomous AI agents. The same coverage said it is available through Antigravity, Gemini API, Gemini Enterprise, Gemini app, and AI Mode in Search.
That wide release puts Gemini 3.5 Flash in front of developers, companies, and regular Google users at once.
Gemini 3.5 Flash is not limited to developers using the Gemini API. The model is now the default option in the Gemini app and AI Mode in Search globally. That makes the release much wider, because Gemini 3.5 Flash can reach people who write code and people who only use Google products for daily questions.
This means the same model family can appear in both consumer tools and developer tools. A regular user may meet Gemini 3.5 Flash while searching for help, planning something, or using the Gemini app. A developer, in another setting, may use it inside an app, an AI agent, or a coding workflow built with the Gemini API.
The release also brings more pressure around safety. Google says Gemini 3.5 has stronger safeguards for cyber risks and chemical, biological, radiological, and nuclear topics. Google also says the model is better at handling sensitive questions, which matters more when AI systems can use tools, write code, and take more steps.
As AI agents become more capable, users will expect more than speed and smart answers. They will need clear controls, visible limits, and human review before important actions are accepted. Gemini 3.5 Flash may help with larger tasks, but trust will depend on how safely and clearly those tasks are handled.
Gemini 3.5 Flash is one of Google’s clearest moves toward agent based AI. It is built for coding, tool use, long tasks, and faster loops. It also connects directly with Antigravity and the Gemini API, which gives developers a practical way to test agent workflows.
Google is placing Gemini closer to the places where real work happens: apps, coding tools, Search, and business systems. Gemini 3.5 Flash fits that direction because it is built for speed, coding, and repeated tasks. Developers will likely test it first in workflows where the model can save time, reduce manual steps, and still leave final control to humans.
Gemini 3.5 Flash is Google’s new AI model for the Gemini API. It is built for coding, AI agents, and long tasks that need several steps. Google says the model is generally available and designed for agent workflows at higher speed and lower cost. It can take text, image, video, audio, and PDF inputs, then return text. Developers can use it through the Gemini API, and regular users may see it in the Gemini app and AI Mode in Search.
Gemini 3.5 Flash is available through several Google products. TechCrunch reported that it is generally available through Antigravity, Gemini API, Gemini Enterprise, Gemini app, and AI Mode in Search. This makes the model useful for both developers and regular users. Developers can build with it through the API, companies can test it through enterprise tools, and consumers can meet it through Google’s own apps. Its reach is wide because Google is placing it across its main AI products.
Gemini 3.5 Flash is useful for coding because it supports tool use, code execution, function calling, structured outputs, file search, and URL context. These features help it act more like a coding agent and less like a simple answer box. It can read files, run commands, test code, and follow several steps. The model is also connected to Antigravity, Google’s agent focused coding environment, which gives developers a more direct place to use it for software work.
Antigravity is Google’s agent focused development environment. It gives Gemini 3.5 Flash a place to plan, write code, manage files, run commands, and use web tools inside a Google hosted Linux sandbox. Google’s Antigravity Agent is available through the Gemini API and Google AI Studio. The idea is to let developers give an agent a task, then let it work through steps in a controlled space. This can help with coding, file work, testing, and research tasks.
Gemini 3.5 Flash is not only for developers. Developers can use it through the Gemini API, Antigravity, and Gemini Enterprise, but Google is also placing it in the Gemini app and AI Mode in Search. That means regular users may use the same model in everyday products. The developer side is more advanced because it includes coding tools and agent workflows. The consumer side is simpler, with Gemini helping inside Google apps and Search experiences.
