Gemini API

Gemini API is Google's LLM API for building apps on Gemini models, with Google AI Studio for prompt testing, a free tier and pay-as-you-go token billing.

Gemini API lets an application send text, images, audio and video to Gemini models and get back text, structured data, speech or images. Google builds and runs it from ai.google.dev, alongside Google AI Studio, the browser workspace where each model can be tried before any code exists. It is a separate product from the Gemini app that consumers chat with. It is also distinct from Gemini Enterprise Agent Platform, the enterprise route to the same models on Google Cloud. Content generation runs on the Flash, Flash-Lite and Pro model lines, and a thinking_level parameter sets how hard a model reasons before it answers. Function calling connects a model to outside tools and APIs, while structured output holds each response to a supplied JSON Schema. Grounding with Google Search and Grounding with Google Maps let answers draw on current web results and place data. Context caching charges less for repeated prompt prefixes, and the Batch API processes many requests in a single call. Beyond text, the API serves Gemini Embedding, the Transcribe speech models, Flash TTS voices, Nano Banana image generation and Gemini Omni Flash for video. It suits developers and product teams putting Gemini inside their own software rather than using it in a chat window. Solo builders and start-ups can prototype on the free tier, where prompts help improve Google products, then switch on paid use once an app meets real users. Paid use is billed by the token for each model, with Batch, Flex and Priority modes for work that can wait or needs priority serving. Rate limits rise through usage tiers as spending grows. Organisations that need contractual data commitments or tuned Gemini models reach the same models through Gemini Enterprise Agent Platform. A typical project starts with a prompt shaped in Google AI Studio and moves into code through the Google GenAI SDKs. Support teams build assistants that answer from current Google Search results, and operations teams use structured output to turn invoices and forms into clean JSON. Travel and local apps add Grounding with Google Maps so answers about places stay current. Search features pair the embedding models with a team's own retrieval store, while nightly classification and summarising jobs move to the Batch API. It fits a developer who wants Gemini behind an API key today and a clear path onto Google Cloud as the project grows.