Gemini: the API with a free tier that is actually usable.

What is Gemini?

Gemini is the name Google uses for its flagship AI models and for the consumer assistant built on them, in the Gemini app, in Android, in Chrome and inside Google Workspace. The models are multimodal from the ground up: they take text, images, audio, video and PDFs as input and produce text (and, in some versions, images and speech). They come in tiers, from a small fast model for high-volume tasks to a large reasoning model, with very long context windows. For developers, the same models are available through the Gemini API and through Google Cloud's Vertex AI.

What is Gemini API?

The Gemini API is the developer interface to those models. You send content (text, images, files) with instructions and receive generated text, structured JSON, function calls or embeddings, with streaming, system instructions, safety settings and a long-context cache. Client libraries exist for Python, JavaScript, Go and others, and there is an OpenAI-compatible endpoint so code written for the OpenAI library works by changing the base URL and key. It is the simplest entry point to Google's models; Vertex AI offers the same models with enterprise features such as data residency and identity-based access.

How to get Gemini API

• Open Google AI Studio and sign in with a Google account.
• Choose Get API key and create one, linked to a new or existing Google Cloud project.
• Copy the key into an environment variable; never into code or a browser app.
• Install the client library in a virtual environment and send a first request with a small model.
• Try the model and prompt in AI Studio's playground first; it shows the code for any prompt you build there.
• Enable billing on the project later if you need higher limits or production use.

What are Gemini API keys?

A Gemini API key is a secret string that authenticates requests to the API and ties usage to a Google Cloud project, where limits and billing apply. Keys created in AI Studio start on the free tier; once billing is enabled on the project, the same key uses paid rates. Treat it like a password: keep it server-side, restrict it in the Cloud console where possible, create one per application, and rotate it if it leaks. Google scans public repositories for exposed keys and may disable them, which is a kindness, but do not rely on it.

Does Gemini Pro have a limit?

Yes, on both sides. In the consumer app, free and paid plans cap how many prompts you can send to the most capable models per period, and the limits differ by plan and change over time. Through the API, every model has rate limits: requests per minute, tokens per minute and requests per day, with the free tier lower than the paid tier, and quotas raised on request for large projects. The context window itself is also a limit, though a very large one. Check the rate-limits page for the current numbers for the model you use rather than relying on any article.

How many Gemini prompts per day

In the Gemini app the daily allowance for the top models depends on your plan and is shown in the app when you approach it; the free plan has a modest daily cap on the strongest model and generous access to the default one. For the API, the free tier sets a requests-per-day limit per model that is enough for development and small projects, and the paid tier raises it substantially. The numbers are revised regularly, so look at the rate-limits documentation for the exact figure today. For a student building a prototype, the free API tier is usually enough until real users arrive.

What is Gemini API cost?

The free tier costs nothing within its limits, with the condition that prompts on the free tier may be used by Google to improve products, so do not send private data there. Paid usage is billed per million tokens, with separate input and output rates per model, discounts for cached context, and higher rates for long prompts on some models. Small models are very cheap; the large reasoning model costs more per token but is competitive with other frontier models. Read the pricing page for current figures, and estimate with your own token counts as I describe in the API cost guide.

What is Gemini API billing?

Billing is handled by the Google Cloud project your key belongs to. Enable billing on that project, attach a payment method, and usage beyond the free tier is charged monthly at the per-token rates. The Cloud console shows usage and cost, and you can set budgets with alerts so a runaway loop cannot surprise you. Projects on the free tier without billing simply stop at the limits rather than charging. The same project can hold keys for several apps; use separate projects when you need to see each app's spend on its own.

How to deploy Google AI Studio app

AI Studio can generate a small web app from a prompt and a model configuration, and it offers a deploy option that publishes the app to Google Cloud Run on your project, giving you a URL. You can also export the code and deploy it anywhere: a Next.js or plain web app can go to Vercel's free plan, a Python backend to Render or Cloud Run. The important detail is the key: a browser app must not embed your API key. Put the model call behind a small server endpoint that reads the key from the environment, and let the page call that endpoint. Cloud Run's free allowance covers a hobby app.

What is Gemini Enterprise?

Gemini Enterprise is Google's business platform for AI agents and assistants across a company's data and applications: pre-built and custom agents, connectors to Workspace and third-party systems, administration, security and governance, sold per user to organisations. It is distinct from the Gemini API, which is the developer interface to the models, and from Gemini in Workspace, the assistant features inside Docs, Gmail and Meet. For a student or a small team the API is the relevant product; Gemini Enterprise is what a company buys when it wants managed agents with enterprise controls rather than building its own.