What is Gemini? A complete map of Google's Gemini ecosystem

What is Gemini? Gemini is the common name Google gives to an entire ecosystem of AI products: a conversational app on web, phone and computer, a family of language models developed by Google DeepMind, a set of APIs for developers, and an AI layer built into Gmail, Docs, Sheets in Google Workspace. Because one word "Gemini" refers to many things, newcomers easily mix them up: paying for Google AI Pro doesn't mean you get free API access. This post doesn't score Gemini as better or worse than ChatGPT, Claude, and doesn't promise any features available in Vietnam for all accounts. Model names, plans and features change very fast; all names and numbers in this post come from the official page on 04/11/2026. The post maps out each segment with links to deeper posts, then wraps up with a table of "use this for that".

One name Gemini, four product layers: app, model, API and Workspace. Diagram: OPCAIOS.

What is Gemini, and why does one name refer to so many things?

Gemini is a common brand, but what you actually use is always one of four specific layers. Telling these four layers apart helps avoid most confusion when reading news, picking a plan or asking support:

1. Gemini app (Gemini application): where regular users chat, at gemini.google.com, Android app, iOS app, and according to official release notes, has a macOS app from 15/04/2026 and Windows from 10/09/2026.
2. Gemini model: the "brain" underneath, like Gemini 4 Argon or Gemini 3.8 Flash.
3. Gemini API: the gateway for software to call the model, priced by token, managed at Google AI Studio.
4. Gemini in Workspace: AI built into Gmail, Docs, Sheets, Meet of a work account, bought by Workspace plan.

Bard is the old name for the chat app; Google changed Bard to Gemini from February 2024, so old posts mentioning "Bard" are talking about what is now the Gemini app in its earlier form.

What is the Gemini app on web and mobile used for?

The Gemini app is Google's chat assistant for personal users, using a Google account, available in free and paid plans: Google AI Plus, Pro, Ultra. According to the help page "Gemini Apps limits & feature availability" (read on 04/11/2026), the free plan still has Canvas, image creation with Nano Banana 2, Deep Research, music creation, slide creation and audio overview; video creation, scheduled tasks and Daily brief need a paid plan. Context window varies by plan: 32,000 tokens on free, 128,000 on Plus, 1 million on Pro and Ultra. Usage limits refresh every 5 hours until hitting the weekly limit. The plan table, pricing and differences are in the post How is free Gemini different from paid plans?.

What do the current Gemini model lines include?

The Gemini model family currently divides by tier: "lightweight" Argon is the top reasoning model, Flash balances speed and cost, Flash-Lite is cheapest and fastest. According to the Gemini API Models page (read on 04/11/2026), the main text models are:

1. Gemini 4 Argon: the flagship reasoning model, 1 million token context, accepts text, images, audio, video and PDF. According to release notes, Argon comes to Google AI Pro and Ultra users from 13/10/2026.
2. Gemini 3.8 Flash: the newest Flash version, aimed at longer-running programming and agents.
3. Gemini 3.6 Flash and Gemini 3.5 Flash-Lite, 3.1 Flash-Lite: stable versions for daily use and high volume.
4. Gemini 2.5 Flash, 2.5 Pro: no longer recommended; the page says they will be switched to new models from 01/12/2026.

There are also specialized models for voice (Live, TTS), transcription, embedding and music (Lyria). If you're just starting, see What is an LLM? to understand language models first.

Gemini API and Google AI Studio, how are they different from the Gemini app?

Gemini API is a service for developers to call the Gemini model from their own software, billed per token, completely separate from the Gemini app subscription. Google AI Studio is a web interface to test prompts, create API keys, and monitor usage. According to the Gemini API Pricing page (read on 04/11/2026), the free tier lets you use some models like Gemini 3.8 Flash and 3.5 Flash-Lite at no cost, but content you send is used to improve Google's products; the paid tier does not. Gemini 4 Argon is not available on the free tier; the price on the page is 2 USD per 1 million input tokens and 10 USD per 1 million output tokens through 30/11/2026, increasing to 4 USD and 20 USD from 01/12/2026. This is the listed price in USD, not including tax. Paying for Google AI Pro does not include an API quota. How to get a key, choose a model, and keep your key safe is covered in Gemini API and AI Studio; for the concept of API see What is an API?.

Gemini in Google Workspace is for whom?

Gemini in Google Workspace is AI built into Gmail, Docs, Sheets, Slides, Meet, and Drive for an organization account, purchased as part of the Workspace plan instead of a personal Google AI plan. According to the Workspace pricing page (read on 04/11/2026, Vietnam region, price in USD per user per month, not including tax), Starter plan is listed at 7 USD, Standard 14 USD, Plus 22 USD; the page currently has promotional pricing for the first 20 users. Gemini in Docs, Sheets, Meet starts from Standard; Starter only has Gmail, Vids, and the Gemini app. The biggest difference is data: the Workspace Privacy Hub states that customer content is not viewed by reviewers and is not used to train models outside the organization without permission. Details for each app are in Gemini in Google Workspace.

Is Gemini CLI still usable?

Gemini CLI is a programming agent that runs in the terminal, open source under Apache 2.0 license, but its role has changed. The official README on GitHub (read on 04/11/2026) clearly states that as of 18/06/2026, the free personal use permission and the permission of Google AI Pro and Ultra plans have moved to Antigravity CLI; Gemini CLI is now for organizations with a Gemini Code Assist Standard, Enterprise, Code Assist for GitHub license, or a paid API key. The README also notes that this tool is not optimized for the Gemini 4 line. Individual users who want a programming agent running Gemini should see What is Antigravity?. Details are in What is Gemini CLI?.

What are Gems, Deep Research, image and video creation, and Gemini Live?

These are features inside the Gemini app, each one solves a different type of task:

1. Gems: a customized assistant with built-in instructions for a repeated task. Release notes from 30/09/2026 say Skills will soon replace Gems and Gems will be migrated automatically. See Gems in Gemini.
2. Deep Research: a self-planning mode that reads many sources and then writes a report. Available in all plans but uses more quota, and free users may not be able to use it during peak times. See Gemini's Deep Research.
3. Image and video creation: images use the Nano Banana family; video is available in paid plans, release notes mention Gemini Omni starting 19/05/2026. See creating images and videos with Gemini.
4. Gemini Live: voice conversation, share camera or screen on mobile. See Gemini Live and voice mode.

Which Gemini product to use for what?

The table below is based on the official page read on 04/11/2026; packages and features may differ by country and account.

What you need	Use	Notes
Q&A, draft writing, summarizing personal documents	Gemini app, free plan	32k token context
Long documents, many files, heavy daily use	Google AI Pro	1 million token context, has Gemini 4 Argon
Market research with sources	Deep Research	Still need to verify sources yourself
Repeating the same task with the same instructions	Gems, moving to Skills soon	Write the instructions once
Illustration images, image editing	Nano Banana	Check for typos, characters
Short video	Paid plan, Google Flow	Uses up quota the most
Integrate into web, chatbot, automated workflow	Gemini API via AI Studio	Charged by token, separate from personal plan
Email, company documents	Gemini in Workspace	Private data terms
Personal coding agent	Antigravity	Gemini CLI now for business
Q&A on your own document set	Gemini Notebook (new name for NotebookLM)	See What is NotebookLM?

If you're unsure between assistants, the post Gemini, ChatGPT or Claude compares them by need.

OPCAIOS experience

Gemini appears in a few specific places in the OPCAIOS system, checked from 03 to 04/11/2026:

1. Orchestration agent: the system uses Google Antigravity, running Gemini models, as an orchestration agent, alongside Claude Code and Codex. The model list on our Antigravity account as of 03/11/2026 includes Gemini 4 Argon and Gemini 3.8 Flash; this list varies by account.
2. Gemini CLI: main machine has Gemini CLI version 0.55.1, extensions not installed, MCP server not configured, productivity not measured.
3. Provider option: admin panel has a "Google Gemini" option for AI writing features, via Google's OpenAI-compatible endpoint.
4. Image creation trial: an internal short video building tool, test run on 03/11/2026, used a Gemini Flash Image model to generate illustration images; results had errors, with repeated images and one wrong character. We haven't used Veo for actual video yet.

We haven't used Gems, Deep Research, Gemini Live, or Gemini in Workspace for evaluation, and don't disclose which paid plans we use.

Frequently Asked Questions

What is Gemini, is it just a chatbot?

It's not just that. Gemini is a collective name for four layers: a chat application, a family of AI models, Gemini API for developers, and an AI layer within Google Workspace. The chatbot that most users encounter is the Gemini application at gemini.google.com or on mobile. The other layers share the same models but have separate pricing, data terms, and purchase options, so when you read Gemini news, check which layer it's about.

Are Bard and Gemini the same?

Bard was the old name for Google's chat application. From February 2024, Google renamed Bard to Gemini, which is the same as the model family name. For this reason, old guides that mention "Bard" often no longer reflect the current interface, plan names, or features. When you encounter old guides, compare them with the Gemini Help page before following them.

Does the Google AI Pro plan give access to the Gemini API?

No. The Google AI Pro plan applies to the Gemini application and some bundled Google services like storage capacity, Google Flow, and Antigravity limits. Gemini API is billed separately by tokens through Google AI Studio, with a limited free tier and a paid tier. According to the Pricing page as of 11/04/2026, content sent through the free tier is used to improve the product, while the paid tier is not.

Is Gemini free in Vietnam?

There is a free plan, you only need a Google account. The official plan page as displayed for Vietnam on 11/04/2026 shows a free plan at 0 đồng, Google AI Plus at 132,000 đồng per month, and Google AI Pro at 489,000 đồng per month; the page does not clearly state whether tax is included. The free plan is limited to 32 thousand tokens of context and lower rate limits. See the plan comparison guide for free and paid tiers in this series for details.

Where should beginners start learning Gemini?

Start with the free Gemini application, pick a recurring weekly task like summarizing emails, and keep note of effective prompts. Once you're comfortable, try Deep Research for a question that needs multiple sources, then verify each source yourself. Only upgrade your plan when you truly hit the limits. If you want structured learning, check out OPCAIOS's AI application training course.

Source: gemini.google/subscriptions and gemini.google/students (Vietnam region), gemini.google/release-notes, support.google.com/gemini (Gemini Apps limits & feature availability, Gemini Apps Privacy Hub, Gemini Live), Models and Pricing pages of Gemini API (ai.google.dev, redirects to aistudio.google.com/docs), README google-gemini/gemini-cli on GitHub, workspace.google.com/pricing, Privacy Hub of Gemini in Workspace (checked on 11/04/2026). Not yet verified: which model the free plan can choose from in the menu on each account, which features are enabled in Vietnam in Vietnamese, when Gems fully switch to Skills. Experience is from OPCAIOS.