Names, versions and owners change fast. Checked October 2026.
In plain words
People often use "ChatGPT" to mean AI in general. It is more specific than that. ChatGPT is an app, a chatbot you type or speak to, on a website or your phone. It was made by the American company OpenAI and launched in November 2022.
The app itself is the window: the chat box, your history, buttons for photos and files. The part that writes the replies is a large language model from OpenAI's GPT family. When OpenAI releases a newer GPT model, the same app gets smarter without changing its name.
The model inside changes as OpenAI releases new versions, so the app's name stays the same while the engine underneath is upgraded. The box above shows which models it runs now. Around the model, the app adds tools such as web search, image making and file reading.
Everyday example
A taxi app and the drivers
You open a ride app, type where you want to go, and a car arrives. The app is the same every day, but the driver who shows up can change. ChatGPT works like that. The app is what you see and tap. The GPT model is the driver doing the actual work, and OpenAI swaps in a newer, better driver every so often.
Where the example breaks: A taxi driver understands the city. A GPT model does not understand the way a person does. It predicts likely words from patterns it learned, which is why it can sound sure and still be wrong.
Latestas of October 2026
- Models inside
- GPT-6 Sol on Plus, Pro, Business and Enterprise plans, and GPT-6 Luna on Free and Go, rolled out from 7 October 2026 (from 8 October on Free and Go), replacing GPT-5.6 Sol and Luna
- New in October 2026
- Intelligent UI: answers can mix text with charts, buttons and other interactive pieces
- Owned by
- OpenAI Group PBC, a public benefit corporation, since 28 October 2025; the non-profit OpenAI Foundation is a major shareholder
These details change often. The rest of this page explains what it is, which stays true.
How it actually works
Behind the chat box, every message goes through the same few steps.
Your message joins the conversation
The app sends the model your new message plus the earlier turns, a hidden set of instructions called a system prompt, and any files or memory you allowed. All of it must fit in the model's context window.
The model writes one token at a time
GPT reads all of that as tokens and predicts the next one, again and again, until the reply is done. This is next-token prediction. Newer models can first spend time on hidden working steps before answering.
Tools fill the gaps
When the model needs fresh facts or a picture, it can call a tool: a web search, a code runner, an image model. The result comes back into the conversation and the model writes the final answer from it. This is tool use.
Trained to be a helpful assistant
A raw language model only continues text. ChatGPT's models are further trained on example conversations and human ratings so they answer questions and follow instructions. That step is RLHF, short for reinforcement learning from human feedback.
- Made by
- OpenAI, San Francisco
- Launched
- 30 November 2022, running a GPT-3.5 model
- What it is
- A chatbot app on the web, desktop and phones
- Earlier models
- GPT-4 (March 2023), GPT-4o (May 2024), GPT-5 (August 2025) and several in between
Other chatbots follow the same pattern: an app on top, a model family underneath. Claude and Gemini are both the app name and the model name.
Questions people ask
Is ChatGPT the same as GPT?
No. ChatGPT is the app you chat with. GPT is the family of large language models made by OpenAI that powers it. Other products, from coding tools to Microsoft's Copilot, have also used GPT models without being ChatGPT.
Who owns ChatGPT?
OpenAI, an American AI company based in San Francisco. OpenAI's business has been restructured more than once since it began as a non-profit lab; the box above shows how it is owned now, and the OpenAI page explains the history.
What does GPT in ChatGPT stand for?
Generative pre-trained transformer. Generative means it creates text. Pre-trained means it first learned from a huge amount of text. Transformer is the design of the neural network inside.
Does ChatGPT search the internet?
It can. On its own the model only knows what was in its training data, up to a cut-off date. When the app's search tool is used, it looks up current web pages and writes an answer from them, usually with links.
Can ChatGPT be wrong?
Yes. It can state false things in a confident tone, a problem called hallucination. Check anything important, such as medical, legal or money matters, against a reliable source.