ToolsCompaniesAbout

Gemini API

by Google@googleMountain View, California, USANo ratings yet

Developer API for Gemini models, agents, and multimodal generation

Visit ai.google.dev
Tool type
Developer API platform for Gemini models and agents
Made for
Developers building applications with Gemini models
Runs on
Remote Gemini API endpoints called from developer code
API calls run on Google-hosted infrastructure; the docs do not describe local runtime. Prototyping can use Google AI Studio.
Last checked
28 Oct 2026
3 product sources

What it does

The Gemini API takes text, images, audio, and video as input, and produces text, images, video, music, speech, and structured reasoning output. Developers call it through the Interactions API, which the docs now describe as the default interface for new projects, while generateContent is treated as legacy. Model examples on the page include a general reasoning model, Flash and Flash-Lite variants, a Live API for streaming voice and video, and image generation models.

Developers use the API in Python, JavaScript, Java, Go, and through raw REST calls, sending a prompt or multimodal input and reading back the output text. Managed Agents run multi-step work in a remote workspace, such as researching topics and producing an HTML dashboard. The Live API supports low-latency voice and video streaming, and Gemini Robotics ER models add spatial reasoning and robot task orchestration.

The API suits developers building prototypes and production applications. A free tier exists for certain models with limited access, while paid usage is metered per million tokens with separate rates for input, output, caching, and batch. Some features, such as the flagship Gemini 4 Argon model, are not available on the free tier. Requests need an API key, and unrestricted keys are rejected.

What it can do

  • Text and multimodal reasoning with Gemini models through the Interactions API
  • Managed Agents that plan and complete multi-step tasks in a remote workspace
  • Live API for real-time voice and video streaming
  • Image generation and conversational image editing (Nano Banana models)
  • Video generation and editing (Gemini Omni models)
  • Music generation (Lyria) and speech-to-text transcription models
  • Gemini Robotics ER models for embodied reasoning and robot task orchestration

Works with

  • Google AI Studio
  • Google Cloud console
  • Files API

Example workflows

Single-prompt text generation
Input: a text prompt such as 'Explain how AI works in a few words'. Steps: create a client, call interactions.create with a model name, read output_text. Output: a short text answer.
Managed agent research task
Input: a multi-step request to research sustainable fashion brands and build a dashboard. Steps: call interactions.create with an agent name and environment set to remote, then wait for the agent to complete the work. Output: the agent's results, which the docs example describes as an analysis.html dashboard.

Before you choose

  • Pricing is per token with separate input, output, caching, and batch rates; prices change on dates shown in the pricing table, such as rate increases starting December 1, 2026 for some models.
  • Gemini 4 Argon is not available on the free tier, and all tokens including thinking tokens are billed.
  • Gemini Robotics ER 1.6 was deprecated on August 31, 2026; new robotics projects should use ER 2 endpoints. Cloud Storage URIs are not accessible through the Developer API, so load images locally or use the Files API.

Listed under

Get started

Keys can be created in Google AI Studio or the Google Cloud console. Unrestricted keys are rejected with a 403 error.

  1. Create a Gemini API key in Google AI Studio or the Google Cloud console (https://ai.google.dev/gemini-api/docs/get-started).
  2. Install the Google GenAI SDK for your language, or prepare REST calls with the x-goog-api-key header.
  3. Start with a single interactions.create call using a model name from the pricing page, then check token usage before scaling.
Create a Gemini API key and call the API

Plans and prices

Prices in USD

Pricing is per 1M tokens and varies by model. Several models have rates that change on December 1, 2026 or January 1, 2027. Batch, Flex, and Priority tiers have different rates, and grounding with Google Search or Maps is billed per 1,000 requests after the free allowance. Free-tier availability differs by model; for example, Gemini 4 Argon is not available on the free tier. Check the pricing page for the current rate table.

FreeUsage-basedFree tier per-token usage on certain models; limits and content-use policies apply.
  • Limited access to certain models
  • Free input and output tokens
PaidUsage-basedMetered by model. See the pricing page rate table for current per-model rates.
  • Higher rate limits for production
  • Context caching and Batch API

Source: ai.google.dev/gemini-api/docs/pricing, checked 28 Oct 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 28 Oct 2026. Each source supports part of the overview; availability and prices can change.

Ratings and reviews

No ratings for Gemini API yet.

Be the first to rate it

What’s changed

Free account

See what’s changed at Gemini API

Launches, price moves and pivots, each with a date and a source. Free with an account.

Signals

Free account

See Gemini API’s signals

Launch traction, hiring and release activity, tracked over time. Free with an account.

Data and security

Free account

See how Gemini API handles your data

Training on your files, compliance, hosting and the models behind it. Free with an account.

Compare with similar tools

Free account

Compare Gemini API side by side

Prices, platforms and features against similar tools, with sources. Free with an account.

Where this comes from

How we research
Company website
ai.google.dev: product, pricing and policy pages
Web search
Confirms the official site
Social accounts
Accounts the company links, and its posts on X
Human review
Checked by a person
knowledge_base

Last checked 28 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.