AI · Windows

Install and Use GPT4All on a Windows PC

Source-checked 9 August 2026About 10 minutes

Answer

Download GPT4All Desktop for Windows from its official quickstart, then select Start Chatting > + Add Model, download a model that fits the PC’s RAM and the intended license, and load it from Chats. GPT4All is a separate Nomic project, not OpenAI’s ChatGPT or a free copy of GPT-4; only a downloaded local model keeps inference on the PC, while an API-connected model sends prompts to its provider.

Check whether the PC can load a local model

Confirm the CPU supports AVX or AVX2 and note the available RAM and storage. In GPT4All’s model list, compare each model’s file size, RAM requirement, and license instead of assuming a larger model will run well.

Install GPT4All Desktop from its official guide

Open the GPT4All Desktop quickstart, click Download for Windows, and run that installer. Complete setup and continue when the GPT4All Home screen appears.

Open the model browser

On Home, click Start Chatting, then + Add Model. The Explore Models screen should open; choose a downloadable local model rather than an API connection if prompts must remain on this PC.

Open GPT4All Desktop quickstart

Download a model that fits the PC

Review the model’s file size, RAM requirement, and license, then click Download. Wait until the completed model appears under Models before opening a chat.

Load the downloaded model in Chats

Select Chats in the left sidebar and click Load Default Model, or choose the downloaded model from the model picker. Wait until its name is shown as loaded before entering a prompt.

Verify local inference

Enter a harmless test prompt and confirm the selected model returns a response. If local-only use matters, disconnect the network and repeat; this test does not apply to an API-connected model.

If it does not work

If a model will not load, compare its listed RAM requirement with available memory and try a smaller supported model instead of downloading from a mirror.
If a model download stalls, check free storage and the Download Path in GPT4All settings, then retry through the in-app catalog.
The installer and model download require internet access. After a local model is downloaded, a network error should not prevent that model from generating a local response.
If responses are very slow, leave Settings > Application > Device on Auto first and try a smaller model before forcing a CPU or GPU option.
If prompts must not leave the PC, load a downloaded local model; GPT4All documents that connected API models send prompts to the remote provider.

Sources and verification

Checked against official documentation on 9 August 2026. Scope: Windows.

Follow the linked first-party guidance if labels differ after a platform or app update.