# ChatGPT how to change model in Work and Codex

> Use the control under the composer in the desktop app or ChatGPT Work, or /model in the Codex CLI, and set a default in config.toml.

URL: https://claudegpt.si/news/chatgpt-how-to-change-model  
Published: 2026-10-29T22:51:13.230Z  
Author: ClaudeGPT Si Newsroom  
Section: Guides  
Topics: ChatGPT, OpenAI, Codex, Codex CLI, How to, GPT-6.1 Sol  
Publisher: ClaudeGPT Si

## Key takeaways

- In ChatGPT Work on the web and the desktop app, the model and reasoning control sits beneath the composer.
- In the Codex CLI, run /model or launch with codex --model gpt-6.1-sol.
- A model = line in config.toml sets the default for the desktop app, CLI and IDE extension.

In ChatGPT Work on the web and in the Codex desktop app, you change the model with the control beneath the message box. In the Codex command line tool you type `/model`. This guide, based on OpenAI's [Models documentation](https://learn.chatgpt.com/docs/models) as of October 29, 2026, covers each place and what the choices mean.

One limit up front: OpenAI's Help Center article on the model picker in regular ChatGPT chat, "GPT-6 and other models in ChatGPT", blocked our requests, so this guide does not describe that picker or the iOS and Android apps. The learn.chatgpt.com page itself sends chat users to that article for those details.

## How to change the model in the desktop app or ChatGPT Work on the web

The steps are the same in both places.

1. Open the app, or open ChatGPT Work in your browser, and start or open a conversation.
2. Find the model and reasoning control beneath the composer, the box where you type.
3. Pick a preset. The default is Power. Move it toward Smarter for deeper reasoning, or toward Faster for quicker, cheaper work.
4. To choose a specific model, open Advanced. There you can set the model, the reasoning effort and the speed.

If Ultra is missing from the slider in the desktop app, go to Settings, then Configuration, and turn on "Ultra in model picker slider".

## Which model to pick

The docs list four models. Names and availability change, so check the picker in your own account.

| Model | What OpenAI says it is for |
| --- | --- |
| Astra (`gpt-6-astra`) | The most capable model, for complex work across code, apps and research |
| GPT-6.1 Sol (`gpt-6.1-sol`) | Close to Astra at a lower cost, for repeated or long-running work |
| GPT-6.1 Terra (`gpt-6.1-terra`) | Fast and efficient, for writing, reading visuals and computer use |
| GPT-6 Luna (`gpt-6-luna`) | The most efficient, for focused, high-volume tasks such as summaries and extraction |

The Power presets named in the docs are Luna High, GPT-6 Sol Light, GPT-6 Sol Medium, Astra Light, Astra Medium and Astra Extra High.

## How to change the model in the Codex command line tool

Run `/model` inside an interactive session and pick from the list. To start a session on a specific model, pass the flag at launch:

```bash
codex --model gpt-6.1-sol
```

`-m` works in place of `--model`, and the flag also works with `codex exec`. We did not run these commands. They come from OpenAI's docs.

## How to set a default model

Add a `model` line to your `config.toml` file:

```toml
model = "gpt-6.1-sol"
```

The desktop app, the command line tool and the IDE extension all read this one file, so the default carries across them.

## What the reasoning levels mean

Reasoning effort controls how long the model thinks before it answers. Higher effort helps on hard tasks but takes longer and uses more tokens, the units models count text in.

- In the desktop app, web and IDE extension the levels are Light, Medium, High and Extra High. The command line tool calls Light "Low".
- Max gives one task more reasoning time. If it is missing in the desktop app, check the app settings.
- Ultra splits work across subagents that run in parallel. Luna supports Max but not Ultra.

OpenAI's advice is to try models on the same inputs and keep the lightest setting that meets your quality bar.

## Why you may not see a model

Availability depends on your plan, the app you use, your workspace settings and the rollout. Per the docs, Free and Go plans get Luna. Plus, Pro, Business, Enterprise and Edu plans can use Sol when it is available. Some paid plans do not include Astra Extra High. On Enterprise and Edu, an administrator may need to turn Luna on first. For current limits and pricing, OpenAI points to its Help Center article on ChatGPT Work and Codex.

Our earlier guide to switching models in a different tool is [Claude Code: how to change the model with /model or --model](https://claudegpt.si/news/claude-code-how-to-change-model).

## Frequently asked questions

### How do I change the model in the Codex CLI?

Run /model in an interactive session, or launch with --model (or -m), for example codex --model gpt-6.1-sol. The flag also works with codex exec.

### How do I set a default model in ChatGPT Codex?

Add a line such as model = "gpt-6.1-sol" to config.toml. The desktop app, the CLI and the IDE extension share that file.

### Why can't I see Astra or Sol in my model list?

Availability depends on your plan, app, workspace settings and rollout. Free and Go plans get Luna, and some paid plans do not include Astra Extra High.

### What is reasoning effort in ChatGPT?

It sets how long the model thinks before answering. The levels are Light, Medium, High and Extra High, and higher levels take longer and use more tokens.

## Sources

- [Models](https://learn.chatgpt.com/docs/models), OpenAI
- [Model selection](https://learn.chatgpt.com/codex/model-selection), OpenAI
- [Models in ChatGPT Work and Codex](https://learn.chatgpt.com/codex/models), OpenAI