Google launches Gemini 3.5 Flash & Antigravity tools

Google has launched Gemini 3.5 Flash and a new set of developer tools centred on its Antigravity platform, extending its push into software agents that can carry out tasks rather than simply respond to prompts.

The announcement includes a standalone Antigravity 2.0 desktop application, Managed Agents in the Gemini API and new Android support in Google AI Studio. Google says Gemini 3.5 Flash outperforms Gemini 3.1 Pro on most benchmarks while running four times faster than other frontier models.

Antigravity is Google's agent-focused development platform, described as a way for developers to turn ideas into production-ready applications. The latest version is intended to give users a central place to work with multiple agents and assign tasks to them in parallel.

The desktop application includes dynamic subagents for parallel workflows, scheduled tasks for background automation and integrations with Google AI Studio, Android and Firebase. Google also introduced an Antigravity command-line interface for developers who prefer terminal-based tools, while Gemini CLI users are being encouraged to move to the new CLI.

Google is also releasing an Antigravity software development kit to give developers programmatic access to the same agent system used in its own products. The SDK lets developers define custom agent behaviour and host agents on infrastructure of their choice.

Antigravity is also being added to the Gemini Enterprise Agent Platform, allowing Google Cloud customers to connect the system directly to Google Cloud projects. That places the new tools within Google's wider cloud and developer software portfolio.

Alongside the product changes, Google introduced a Google AI Ultra subscription priced at USD $100 a month. The plan offers five times the usage limit in Google Antigravity compared with Google AI Pro, and includes USD $100 in bonus Antigravity credits for new and existing subscribers who reach their quota limit.

Managed agents

A separate part of the launch focuses on the Gemini API, where Google is introducing Managed Agents. With a single API call, developers can create an agent that reasons, uses tools and executes code in an isolated Linux environment, according to Google.

These managed agents use the Antigravity agent harness and are built on Gemini 3.5 Flash. They are available through the Interactions API and in Google AI Studio.

Each interaction creates a persistent isolated environment that can be resumed in later calls with files and state intact. That is intended to support multi-turn sessions in which an agent continues work over time rather than starting from scratch with each request.

Developers will also be able to define their own agents by extending the Antigravity system with custom instructions and skills through markdown files. New custom agent templates are also being added to the Google AI Studio Playground.

Studio expansion

Google AI Studio is also expanding across mobile and other parts of Google's software ecosystem. A mobile app for Google AI Studio is available for pre-registration and is intended to let developers capture ideas away from their desks and return to a working prototype later.

Workspace integration is another part of the update. Agents will now be able to call relevant Google Workspace application programming interfaces directly and embed them into applications.

Google has also added an export function from Google AI Studio into Antigravity. This will allow developers to move whole projects into local development and then into production while retaining project context.

For Android developers, AI Studio now includes native support that lets users build Android applications from a prompt. It is also adding support for Google Play Console so developers can publish apps directly to the test track from AI Studio.

Competition launch

Google also announced the Build with Gemini XPRIZE Hackathon, a global competition with a USD $2 million prize pool. Google says this is the largest prize pool it has offered for a hackathon.

The competition is aimed at developers using Gemini models, tools and platforms to create applications focused on issues including food waste and scientific research. Participants will be building applications rather than demonstrations.

The launches show how major technology groups are trying to move artificial intelligence products beyond chat interfaces and into software that can take actions on behalf of users. In Google's case, that strategy now spans model development, application interfaces, desktop software, mobile tools and cloud integrations.

"The leap from single-turn prompts to collaborative, always-on agents changes how developers build software - and we're just getting started. Our goal is to make that shift easier for you. With the new Antigravity platform, Managed Agents in the Gemini API and expansions to Google AI Studio, we are removing infrastructure friction so you can focus on bringing your ideas to life," Google said.