Understanding AI

Google AI Studio: how-to guide to explore and prototype with Gemini

Google AI Studio is Google's web-based environment for testing Gemini models and prototyping generative AI applications, with no infrastructure to set up. Access is free with a Google account. Whether you are a developer or a marketer, it…

Jean-Baptiste Duquesne

Jean-Baptiste DuquesneSEO, GEO & content operations · Published 06 May 2025 · Updated 27 October 2026

Illustration for the article: Google AI Studio: how-to guide to explore and prototype with Gemini

Google AI Studio is Google's web-based environment for testing Gemini models and prototyping generative AI applications, with no infrastructure to set up. Access is free with a Google account. Whether you are a developer or a marketer, it is the fastest way to see what the latest models can do on your own use cases, and how smart automation can support your work.

Updated in October 2026: available models, Build mode, free-tier conditions and moving to production.

Which Gemini models are in AI Studio in October 2026?

The line-up has changed a lot since the Gemini 2.5 generation. The main models available are:

  • Gemini 4 Argon, generally available since October 13, 2026: Google's most capable model, built for complex coding, agentic, math and science tasks.
  • Gemini 3.8 Flash: the most capable Flash model, designed for agents and enterprise workflows, with a good balance of quality, speed and cost.
  • Gemini 3.5 Flash-Lite: the lightest and most cost-efficient option for high-volume tasks.
  • Specialized models: image generation and editing (the Nano Banana family), real-time voice conversation (Gemini 3.8 Live), text-to-speech and transcription.

Since September 2026, Gemini 2.5 models are no longer offered to new users: Google recommends Gemini 3.8 Flash or 3.5 Flash-Lite for any new project.

An interface to test, then to build

In the chat workspace, you write a prompt, add documents, images, audio or video, and compare model responses. You can turn on tools such as Google Search grounding to base answers on up-to-date sources, or request structured output (JSON) to feed another tool. It is a hands-on way to practise prompt engineering and to explore AI-assisted content creation.

Available settings depend on the model: in July 2026 Google deprecated the classic sampling parameters (temperature, top_p, top_k) on its recent models. The Get Code feature still turns a working prompt into Python, JavaScript or cURL code for the Gemini API.

The biggest addition is Build mode: you describe an app in plain language and AI Studio generates it, either as a full-stack web app (React front end and Node.js server) or as an Android app. You can then deploy it to Cloud Run, sync it with a GitHub repository or export the code.

Use cases: exploring Gemini's capabilities

Google AI Studio is an ideal playground for discovering how multimodal AI can transform your workflow. From AI for marketing automation to deep data synthesis, the possibilities are vast. You can generate specialized copy, summarize internal documents, analyze images and video, or perform AI clustering by feeding the model unstructured data to find patterns.

For strategic planning, the tool supports in-depth research by letting you query models for market insights or competitor analysis. It is also a testing ground to compare Gemini with other model families, such as those from OpenAI or Anthropic, and choose the right tool for the job. Agentic experiments are another useful reference: Anthropic's work on Claude in Minecraft shows how a model can plan and build inside a game environment through text commands.

Limitations, privacy and scaling up

The free tier covers the Flash, Flash-Lite, Live and text-to-speech models, within usage limits. Gemini 4 Argon and the image generation models are not included in the free tier. A key point for businesses: on the free tier, content you submit may be used by Google to improve its products, which is not the case on the paid tier. Avoid testing confidential data without moving to a paid plan; this is part of responsible AI use and of treating AI adoption as an organizational challenge rather than just a technical one.

For enterprise-scale deployment, with data governance, security and agent management, Google now points businesses to Gemini Enterprise Agent Platform, which took over from Vertex AI in April 2026. To understand how models are adapted to proprietary data, check out What is SFT? for a detailed explanation of supervised fine-tuning.

Brandeploy: scale your Gemini prototypes into global brand excellence

Brandeploy is a brand management and creative automation platform that helps enterprise teams scale content production while maintaining strict consistency. By taking the winning prompts and content structures discovered in Google AI Studio, teams can use Brandeploy to automate the generation of brand-compliant assets across hundreds of markets. The platform ensures that the creative power of Gemini is harnessed within your brand guidelines, preventing unauthorized variations and centralizing your creative operations. Discover how Brandeploy's AI creative automation produces on-brand banners, videos and visuals at scale. To see how we can transform your AI-driven production, book a demo.

FAQ

What is the difference between Google AI Studio and Vertex AI?

Google AI Studio is a fast, web-based environment to test and prototype with Gemini models. For enterprise-scale production (data governance, security, agent management), Google now offers Gemini Enterprise Agent Platform, which took over from Vertex AI in April 2026.

Is Google AI Studio free to use?

Yes, access to AI Studio is free, and the API free tier covers the Flash, Flash-Lite, Live and text-to-speech models within usage limits. Gemini 4 Argon and image models are paid. On the free tier, submitted content may be used to improve Google's products; it is not on the paid tier, which is the better choice for business data.

Which Gemini models are available in AI Studio?

As of October 2026: Gemini 4 Argon, the most capable model (generally available since October 13), Gemini 3.8 Flash for balanced quality, speed and cost, Gemini 3.5 Flash-Lite for high volumes, plus image (Nano Banana), real-time voice (Live), text-to-speech and transcription models. Gemini 2.5 models are no longer offered to new users.

How can I export my work from AI Studio to a real application?

Use the Get Code feature to turn a refined prompt into Python, JavaScript, or cURL code for the Gemini API. With Build mode, you can go further: generate a full app from a description, deploy it to Cloud Run, sync it with GitHub or export the code.

What is Build mode in Google AI Studio?

Build mode lets you create an app by describing it in plain language. AI Studio generates a full-stack web app (React and Node.js) or an Android app, which you can then refine by prompting, deploy to Cloud Run, sync with a GitHub repository or download as code.

Jean-Baptiste Duquesne

Written by

Jean-Baptiste Duquesne

SEO, GEO & content operations

Web pioneer since 1995, founder of the recipe site 750g.com (sold to Webedia), now industrialising SEO and GEO content production with AI.

Ready to scale your content production?

Book a demo