# Is ChatGPT an LLM, and which model does GitHub Copilot use?

Source: https://codelifter.com/forum/posts/is-chatgpt-an-llm-and-which-model-does-github-copilot-use/
Site: CodeLifter Forum, category Vibe coding prompts
Published: 2026-11-10. Updated: 2026-11-10. Replies: 1.

## Question by CodeLifter editors (staff), 2026-11-10

ChatGPT is not itself an LLM: it is OpenAI's chat app, built on OpenAI's GPT large language models. GitHub Copilot does not use one model either. It offers models from OpenAI, Anthropic, Google, xAI, Microsoft and Moonshot AI, chosen by you or by its Auto setting, and inline suggestions run on a separate completion model.

## Is ChatGPT an LLM or an app built on one?

An LLM, a large language model, is the trained model that predicts the next piece of text; the [LLM thread](/forum/posts/what-is-an-llm-and-how-does-it-write-code/) explains how. ChatGPT is the product around one: the website and apps with a chat window and your conversation history. People say "ChatGPT is an LLM" because the chat is the only part they see, but OpenAI's own documents keep the two apart.

OpenAI's help article [What is ChatGPT?](https://help.openai.com/en/articles/6783457-what-is-chatgpt) says "ChatGPT is fine-tuned from GPT-3.5, a language model trained to produce text", optimised for dialogue with reinforcement learning from human feedback. That article describes the original release; some of its other answers (for example a knowledge limit of 2021) are out of date, so read it as history. The current picture is on OpenAI's [models page](https://developers.openai.com/api/docs/models), read on 16 October 2026, which lists API models labelled as the ones used in ChatGPT, such as "GPT-5.3 Instant model used in ChatGPT" and `chat-latest`, which "points to the latest Instant model currently used in ChatGPT."

So the accurate sentence is: ChatGPT runs on GPT models, and which one answers can change as OpenAI updates the app.

## Which model does GitHub Copilot use?

It depends on the plan, the feature and your choice. GitHub's [supported models page](https://docs.github.com/en/copilot/reference/ai-models/supported-models), read on 16 October 2026, lists 32 generally available models:

| Provider | Models listed | Examples |
| --- | --- | --- |
| OpenAI | 13 | GPT-5 mini, GPT-5.3-Codex, GPT-5.5, GPT-6 Astra |
| Anthropic | 11 | Claude Haiku 4.5, Claude Sonnet 5.5, Claude Opus 5.5 |
| Google | 3 | Gemini 3.8 Flash, Gemini 4 Argon |
| xAI | 3 | Grok 4.5, Grok 4.6, Grok 4.7 |
| Microsoft | 1 | MAI-Code-1.1-Flash |
| Moonshot AI | 1 | Kimi K3 |

The page notes that "Model availability is subject to change", so treat the counts as a snapshot. Which ones you can use depends on where you are:

- **Chat, agent mode, Copilot CLI and the cloud agent.** You pick a model, or choose **Auto**. GitHub's [auto model selection page](https://docs.github.com/en/copilot/concepts/auto-model-selection) says Auto combines a system that "tracks real-time system health and availability" with one that "evaluates task complexity", then routes each task to a model. A simple prompt can go to a small model and a hard one to a larger model.
- **Copilot Free and Copilot Student.** The [plans page](https://docs.github.com/en/copilot/get-started/plans) says models on these plans are available "through auto model selection only", so you cannot pick one.
- **Inline suggestions.** The grey ghost text uses its own completion model. GitHub's guide to [changing the completion model](https://docs.github.com/en/copilot/how-tos/copilot-in-your-ide/copilot-for-common-tasks/change-the-completion-model) uses the command **GitHub Copilot: Change Completions Model**, and notes that when only one model is available, the picker shows only that one.
- **Business and Enterprise seats.** An administrator can enable or restrict models for the organisation, and Auto will not use a model your policies exclude.
- **Fallback.** The supported models page says that "as the long term support model, GPT-5.3-Codex will be used in the event no other models are available."

## How do you find out which model answered?

The auto model selection page gives the method for each surface:

1. In Copilot Chat and the GitHub Copilot app, hover over the response.
2. In Copilot CLI, the model is shown in the terminal for each response.
3. In the cloud agent, it appears at the end of the response.
4. For inline suggestions in VS Code, open Settings and search `copilot completion`. If the "Selected Completion Model" field is empty, the default model is in use.

## Does the model matter for your code?

Yes, but no source we can cite ranks them for your project. GitHub's own [model comparison](https://docs.github.com/en/copilot/reference/ai-models/model-comparison) recommends models by task area (general coding, quick simple tasks, deep reasoning and debugging, working with visuals); that is guidance, not a test on your code, and the scores on the CodeLifter [models page](https://codelifter.com/models/) rate chat answers, not how a model edits a real repository. Two things are worth knowing. First, cost differs: GitHub's billing pages say credit use depends on the model, and paid plans get a 10% discount on model costs when using Auto. Second, Auto may send some prompts to evaluation models, covered in the follow-up below.

If you are starting out, Auto is a sensible default; switch to a named model when a task keeps failing and compare the results on the same prompt. The [first-time Copilot thread](/forum/posts/how-do-you-use-github-copilot-for-the-first-time/) covers setup, and the [coding agents thread](/forum/posts/claude-code-codex-or-copilot-how-do-coding-agents-differ/) compares Copilot with Claude Code and Codex.

## Short version

- ChatGPT is an app; the LLMs underneath are OpenAI's GPT models, and they change over time.
- Copilot offers models from six companies; you choose one or let Auto route each task.
- Free and Student plans get Auto only; inline suggestions use a separate completion model.
- To see which model answered, hover over a chat response or check the CLI output.

## Reply 1 by CodeLifter editors (staff), 2026-11-10

**What is the difference between GPT and ChatGPT?** GPT is the name of OpenAI's family of language models; ChatGPT is the consumer app that uses them. You reach a GPT model directly through OpenAI's API, where you name the model in each request. OpenAI's [models page](https://developers.openai.com/api/docs/models) (read 16 October 2026) points developers to GPT-6 Astra as "our flagship model for complex reasoning and coding", and separately lists models labelled as the ones used in ChatGPT, such as `chat-latest`. The `chat-latest` page says it points to the latest Instant model in ChatGPT, that "The underlying model snapshot will be regularly updated", and recommends GPT-6 Astra for production API use instead. That is the practical difference for a developer: in ChatGPT the app decides the model and may change it without notice, while in the API you pin a model name and get repeatable behaviour until that model is deprecated. Data handling differs too. OpenAI's [What is ChatGPT?](https://help.openai.com/en/articles/6783457-what-is-chatgpt) article says that for individual services "we may use your content to train our models" unless you opt out, and that business offerings such as ChatGPT Enterprise are not used for training.

---
Cite as: CodeLifter Forum, "Is ChatGPT an LLM, and which model does GitHub Copilot use?", https://codelifter.com/forum/posts/is-chatgpt-an-llm-and-which-model-does-github-copilot-use/, accessed 2026-11-12.