AI models for coding, inside your editor

Open the DEVUP AI panel and reach the model picker from the message box. Type to search it, narrow it down with the capability filters, or scan the models grouped under each maker, every one shown with its maker's logo.

Whatever you choose is still selected the next time you open the editor, so you only pick again when you want a change.

1. 1In the DEVUP AI panel, click the model shown in the message box to open the picker.
2. 2Type part of a name, or narrow the list with a capability filter. Models are grouped by model maker, each group with its maker's logo.
3. 3Pick a model and attach the code you plan to refactor, then describe what should change.
4. 4Read the answer. The reply names the model that wrote it, and shows how long it took and its token count.
5. 5Not convinced? Get that last reply once more from a different model, without touching your default.
6. 6When you know which model suits quick jobs, enter its id in devupai.editorActions.model so editor actions use it while the chat keeps its own.

A conversation is not locked to one model. Change it from the message box at any point, and each reply stays labelled with the model that wrote it.

Curious how a different model would answer? Have another model write the last reply again; your default model stays as it was.

Comparing answers before you commit to one. Ask a design question, then have a second model produce the last reply again. Each reply carries the name of the model behind it, so you always know whose suggestion you are reading.

Following a reasoning model's work. For a tricky algorithm, choose a model that reasons before it answers. A Thinking indicator counts the seconds until the reply begins, you can follow the thinking as it happens, and it can be reopened under the reply later.

Sending a large file. The gauge estimates how much of the selected model's context your message would use. If it is too big for that model, DEVUP AI keeps it from going out, explains the reason and lets you pick another model.

Generating a long file. When a reply stops at its length limit before the code is complete, continue it in place. For models you use for long output, raise the longest reply in Settings, within each model's own limit.

• Editor actions: Explain, Fix, Add Tests and Quick Fix can run on a model of their own, or follow whichever model the chat uses. See editor actions.
• Reply length: cap how long each model's replies can be, up to that model's own limit.
• Reasoning models: watch a reasoning model think while it works, and open that thinking again under the reply afterwards.
• Logos: model logos can be switched off in the appearance settings if you prefer a plainer list.

Installing the extension costs nothing. The requests you send to any model are paid from your DEVUP balance, which is kept in Algerian dinars and topped up with BaridiMob, CIB or Edahabia.

Your balance is visible in the status bar and at the top of the panel. If it runs too low for a request, DEVUP AI tells you and offers Top up and Refresh balance.

Each model in the list below opens its own page, and each family name opens the family's page. To look beyond the editor, browse the DEVUP AI model catalog or the model families.

The chat and models docs walk through the picker, switching and History step by step.