What is GPT?
GPT is the name of OpenAI's family of language models, the engines behind ChatGPT and many other apps. Here is what the letters mean, how the models work and which versions matter today.

GPT in one paragraph
GPT is a family of large language models made by the American company OpenAI. A language model is a program trained on an enormous amount of text so that it can continue, answer, translate, summarize and rewrite text on request. GPT models are the engines inside ChatGPT, inside thousands of apps built on OpenAI's API, and inside independent sites like this one. When people say "ask GPT", they usually mean typing a question into one of those apps and getting a written answer back a few seconds later.
The newest generation, released between September and October 2026, is called GPT-6. Instead of one model it is a small family with four names: Astra, Sol, Terra and Luna. Further down you will find what each one is for and which two of them you can use for free in our chat.
What the letters G, P and T stand for
The name comes from a 2018 research paper and describes how the models are built:
- Generative. The model produces new text rather than picking an answer from a list. Every reply is composed on the spot, which is why asking the same question twice can give two differently worded answers.
- Pre-trained. Before it learns to chat, the model goes through a long first stage of training on a huge collection of books, articles, code and web pages. That stage gives it general knowledge of language and of the world. Later stages shape it into a helpful assistant.
- Transformer. This is the type of neural network underneath, introduced by Google researchers in 2017. A transformer is good at working out which earlier words matter for the next one, even when they are far apart in a long text.
So "GPT" literally means a text generator, trained in advance, built on a transformer network. The number after it (GPT-4, GPT-5, GPT-6) marks the generation. Since mid-2026 OpenAI also adds a tier name such as Sol or Luna to tell bigger and smaller models of the same generation apart.
How a GPT model writes an answer
Under the hood a GPT model does one thing very well: given some text, it predicts what piece of text should come next. Text is split into tokens, small chunks that are often a whole short word or part of a longer one. The model reads your message as tokens, calculates which token is most fitting next, adds it, and repeats until the answer is complete. That is why answers appear word by word on the screen.
Predicting the next token sounds simple, but doing it well across billions of examples forces the model to pick up grammar, facts, styles, reasoning patterns and even some common sense. Modern GPT models go through several training stages:
- Pre-training on a very large text collection, which teaches language and broad knowledge.
- Instruction tuning on examples of good answers, so the model follows requests instead of just continuing text.
- Feedback training, where people and automated graders rate answers and the model learns to prefer helpful, honest and safe ones.
- Reasoning training, used since OpenAI's o1 models in 2024, which teaches the model to work through a problem step by step before it writes the final reply.
That last stage is why recent models have a "thinking" setting. With little or no thinking the reply starts almost at once; with more thinking the model spends extra time, and extra tokens, checking its work. In our chat, the Think chip switches from the quick GPT-6 Luna to GPT-6.1 Terra, which thinks briefly first.
Two limits follow from how all this works. First, a model only knows what was in its training data, up to a knowledge cut-off date. Second, it writes what is likely, not what it has verified, so it can produce confident mistakes. Both come up again below.
Who makes GPT
GPT models are developed by OpenAI, the San Francisco company founded in 2015 that launched ChatGPT in November 2022. OpenAI sells access to its models in two ways: through ChatGPT, its own assistant with free and paid plans, and through its API for developers, who pay per token. Big cloud providers, including Microsoft Azure and Amazon Bedrock, also offer recent GPT models to their business customers.
Because the API is open to any developer, many other products run GPT models too. Our site is one of them. GPT.me is an independent website, not affiliated with, endorsed by or run by OpenAI. We connect to OpenAI's models through a cloud API, and we do not train or host them ourselves. "GPT" and "ChatGPT" are trademarks of OpenAI.
The GPT-6 family in November 2026
OpenAI introduced its current naming with GPT-5.6 in July 2026: the number is the generation, and Sol, Terra and Luna are tiers that can be updated on their own schedule. GPT-6 added Astra on top. Here is the line-up today:
| Model | Released | Role in the family | In our chat |
|---|---|---|---|
| GPT-6 Astra | September 3, 2026 | Flagship, for the hardest coding, science and professional work | No |
| GPT-6.1 Sol | September 29, 2026 | High-end model, an upgrade to GPT-6 Sol | No |
| GPT-6 Sol | September 22, 2026 | High-end model; runs ChatGPT for paying users | In our AI Chat app |
| GPT-6.1 Terra | October 26, 2026 | Mid-tier: images, natural writing, computer use | Yes, with Think |
| GPT-6 Luna | September 22, 2026 | Fastest and cheapest; runs ChatGPT's free plan | Yes, default |
All five accept text and images and reply in text, and all have a context window of about 1,050,000 tokens on OpenAI's platform. They differ in depth, speed and price. You can read more on the pages for GPT-6 Luna, GPT-6.1 Terra and GPT-6 Sol, or follow the whole story in our GPT history.
GPT, ChatGPT and GPT.me: not the same thing
These names get mixed up all the time, so it helps to separate them:
- GPT is the model: the trained network that turns your words into an answer.
- ChatGPT is OpenAI's product built around those models. It adds an account, saved history, web search, voice, picture creation, file uploads and many other tools.
- GPT.me is our free website. It sends your message to a real GPT model (GPT-6 Luna, or GPT-6.1 Terra with Think) and shows the answer in a simple chat. It has no account, keeps your history only in your browser, and has none of ChatGPT's extra tools: no web search, no voice, no image creation and no memory between chats.
Our GPT.me vs ChatGPT page lists the differences in detail, including the many things ChatGPT does that our site does not.
What GPT does well, and where it slips
A current GPT model is a strong general helper. It is good at:
- Writing and editing: drafts, emails, summaries, rewrites in another tone. Try it: rewrite a message.
- Explaining: a hard concept at the level you need. Example: how vaccines work.
- Coding: reading errors, writing small functions, explaining unfamiliar code.
- Images you send: describing a photo, reading a chart, transcribing a handwritten note.
- Languages: it understands and writes dozens of languages and translates between them.
It also has real weak spots. It can invent facts, quotes and sources that look convincing. It does not know about events after its cut-off date unless a tool feeds it fresh information, and our chat has no such tool. It can miscount, misread a chart or slip on long calculations. It is not a doctor, lawyer or financial adviser. Treat its answers as a capable first draft and check anything important; our guide to fact-checking AI answers shows simple ways to do that.
How to try GPT for free
You have several options, each with its own trade-off. OpenAI's ChatGPT has a free plan that runs GPT-6 Luna. Some apps, including our own AI Chat app, include GPT models. And this site gives you GPT-6 Luna and GPT-6.1 Terra in the browser, with no sign-up and a few free messages every day.
To start, open the chat and type a question, or pick one of the ready-made AI tools such as the summarizer or the grammar checker. If you are new to writing prompts, our guide to GPT prompts that work takes five minutes and makes a noticeable difference. A complete overview of the free routes is on how to use GPT for free.
FAQ
What does GPT stand for?
Generative Pre-trained Transformer: a model that generates text, is trained in advance on a large amount of writing, and uses the transformer type of neural network.
Is GPT the same as ChatGPT?
No. GPT is the family of models; ChatGPT is OpenAI's app that uses them and adds tools such as web search and voice. Other apps and sites, including ours, use GPT models too.
What is the latest GPT model?
As of November 2026 the newest release is GPT-6.1 Terra, from October 26. The most capable model is GPT-6 Astra, OpenAI's flagship since September 3.
Is GPT.me made by OpenAI?
No. GPT.me is an independent site. It uses real OpenAI GPT models through a cloud API, but OpenAI does not run or endorse it.
Does GPT know about today's news?
Not on its own. GPT-6 Luna's knowledge stops in May 2026 and GPT-6.1 Terra's in June 2026. ChatGPT can search the web to fill the gap; our chat cannot, so check anything recent yourself.
Got a question? Ask it now
No account and no download. Open the chat, type what is on your mind, and the answer starts appearing in a second or two.
Start chatting
