What is Gemini?

Gemini is Google's family of large language models and the consumer app built on top of them. It replaced Bard in 2024 and is made by Google DeepMind. When people say "Gemini" they usually mean one of three things: the chat app at gemini.google.com, the models inside Google Workspace (Gmail, Docs, Sheets), or the Gemini API that developers call.

Under the hood, Gemini is a multimodal model. That means you can paste text, drop in a PDF, upload a photo or a video, and ask questions about all of it in the same prompt. Most rivals bolt vision on afterwards; Gemini was trained on mixed inputs from the start. For a beginner the practical result is simple: you rarely have to convert a file before handing it over.

Three ways to use it, three price tags

• The Gemini app (free). Sign in with a Google account and chat. Rate limits apply and the strongest models may be restricted. Costs $0.
• A Google subscription. Google AI Pro ($19.99 a month in our catalogue at the time of writing) and Google AI Ultra (listed at $249 a month) unlock higher limits, the top models, Gemini inside Workspace and a storage bundle. See Google AI plans.
• The Gemini API. Pay per token via Google AI Studio or Vertex AI. There is a free tier for testing. This is where the cheapest prices live if you use Gemini at volume — lesson 7 covers it.

What Gemini is good at

Three things stand out. First, long context: current Gemini API models list a context window of 1,048,576 tokens in our catalogue, which is roughly 700,000 words. You can drop a whole codebase or a year of meeting notes into one prompt. Second, multimodal input: screenshots, PDFs, audio and video all work without pre-processing. Third, Google integration: with a paid plan Gemini reads your Gmail, Drive and Calendar when you allow it, and it lives inside Docs and Sheets.

Google also releases Gemma, a set of open-weight models you can run yourself or rent cheaply from hosts. They are not Gemini, but they share lineage and appear under Google in our catalogue. If you are cost-obsessed and technical, Gemma is worth a look in lesson 2.

Honest limits

• It is confidently wrong sometimes. Like every LLM, Gemini hallucinates. Grounding with Google Search helps but does not remove the problem. Verify anything that matters.
• The product surface moves fast. Model names, plan names and included features change often. Google One AI Premium became Google AI Pro; Gemini Advanced became a feature rather than a plan. Check the official plans page before relying on a specific feature.
• Free tier limits are not published in detail. You will hit a wall on heavy use and Google does not say exactly where. Budget on the assumption that the free tier is for light, occasional work.
• Privacy settings matter. Gemini Apps Activity is on by default for many accounts. Lesson 4 shows where to change it.