Google's AI tools sorted: which ones you need and which can wait

Google has so many AI products that the video above counts more than thirty. Some are named after the model (Gemini), some after a code name (Nano Banana), and some exist only as experiments on labs.google. You don't need to know them all. For most people a handful really matter, and the rest can wait until you have a concrete reason. Here is which ones, and what you use them for.

As of September 2026. This article first appeared in January 2026 and has been updated. Google changes names, plans and availability quickly; Labs experiments are not always available outside the US.

Gemini: the basis for text, research and long documents

Gemini is Google's chatbot and also the name of the model behind almost every other tool. For everyday use it is comparable to ChatGPT. Two things set it apart.

Deep Research. You ask a research question, Gemini drafts a plan, searches the web and delivers a report with sources. It takes minutes instead of seconds, but you get a worked-out piece rather than an answer. Deep Research is included in every plan, including the free one.

The context window. On a paid plan Gemini can process a lot of text at once: 1 million tokens in Google AI Pro and Ultra. Google puts that at roughly 1,500 pages of text or 30,000 lines of code. The free version has 32,000 tokens, Google AI Plus 128,000. The paid plan formerly called Gemini Advanced has been called Google AI Pro since May 2025.

That large window helps when you want one long contract or an entire codebase read in one go. For a question about an email it makes no difference.

NotebookLM: when the answer has to come from your documents

NotebookLM (Gemini Notebook since July 2026) works with sources you provide: PDFs, Google Docs, websites, videos. Answers include a citation to the passage they came from, so you can check them. From the same sources it makes podcasts (Audio Overviews), explainer videos, mind maps, tables and quizzes.

The difference from Gemini: Gemini gets information wherever it can find it; NotebookLM sticks to your sources. That doesn't stop an answer from being wrong, but you can see where it came from. This article on the big NotebookLM update lists the features and limits.

Images and video: Nano Banana and Veo

Nano Banana is the nickname of Google's image model. The first version (Gemini 2.5 Flash Image, August 2025) became known for keeping the same character or product recognisable across different scenes. That was a well-known weak spot of image generators: ask for the same person in a different setting and you often got a different face. There is now Nano Banana 2, available to everyone in the Gemini app, and Nano Banana Pro for paying subscribers.

Veo is the video model. Since Veo 3 (May 2025) it generates picture and sound in one pass: dialogue, ambient sound and effects, in sync with what you see. The current version is Veo 3.1. Video generation in the Gemini app requires a paid plan.

If you want to use these models outside Google, for example in your own application or a design workflow, platforms such as fal.ai (via an API) and Freepik (recently renamed Magnific) offer the Nano Banana models.

Gems and Opal: building your own assistant

Gems are assistants in Gemini with fixed instructions and optionally your own files. You write down once what a proposal or quarterly plan should look like, and you don't have to repeat it in every conversation. Nothing is trained: a Gem is a saved instruction with context.

Opal goes a step further. It is a Google Labs experiment for building small AI apps without code, chaining prompts, models and tools together. You describe what the app should do, Opal turns it into a visual flow, and you adjust it in plain language or in the editor. The original version of this article compared Opal to Zapier and Make, but that's only half right: Zapier mainly connects existing apps, while Opal builds an app around an AI model. Opal is free, available in more than 160 countries and also sits in the Gems manager of the Gemini web app.

Google Labs: where the experiments live

On labs.google Google puts tools that are still being tested. Some become products, others disappear. Three that come up in the video:

• Pomelli analyses your website and builds a "Business DNA" from it: tone of voice, colours, fonts and visual style. It then generates campaign material for social media and ads in your own brand style. Built by Google Labs and DeepMind, launched in October 2025.
• Disco is an experimental browser. Its main feature, GenTabs, turns your open tabs and chat history into a small interactive app, for example a trip planner built from ten tabs of hotels and train times. Announced in December 2025, with a waitlist and initially macOS only.
• Mixboard is an AI mood board: start from a description or your own images and refine with instructions like "make it more minimal". It runs on Nano Banana.

Keep in mind that Labs experiments often launch in the United States or a few other countries first.

Where to start

My advice: start at labs.google to see what's coming. Make sure you master NotebookLM for everything involving your own documents, and use Deep Research in Gemini for questions that would normally cost you an afternoon of searching. Images and video can wait until you have a concrete job for them.

Google's ecosystem is a big toolbox. You don't need to know every screw, as long as you know which machine does the heavy lifting.