How to install OpenAI GPT-OSS AI model on Windows 11

To install the GPT-OSS AI model on Windows 11, first, open Command Prompt (admin) and run the winget install --id Ollama.Ollama command.
Second, you need to select the gpt-oss:20b option from the list and submit a question to trigger the download.
Alternatively, you can use the Ollama command-line tool with Command Prompt (admin), and use the ollama pull gpt-oss:20b command to download and install (for example) the GPT-OSS locally on your computer.

On Windows 11, you can install and run the latest GPT-OSS (Open Source Software) AI model from OpenAI locally on your computer, and in this guide, I’ll outline the steps to complete this configuration.

GPT-OSS is a text-based model from OpenAI’s family of open-weight large language models. Currently, we have two versions, including gpt-oss-20b (with around 21B parameters) and gpt-oss-120b (with around 117B parameters), and they have been released under the Apache 2.0 license, allowing anyone to download, run, modify, and redistribute them freely.

gpt-oss-20b: Compact enough to run on regular hardware with a minimum of 16GB of memory (for example, laptops, powerful phones) and excels at local, low-latency tasks.
gpt-oss-120b: It’s the larger model, designed for powerful reasoning tasks, and it runs on high-end setups like a single 80GB GPU and offers near-O4-mini performance.

These models support chain-of-thought reasoning, tool use, long context windows, and offer configurable reasoning levels, such as low, medium, and high, for flexibility.

What can I do with GPT-OSS as a regular user?

If you’re a tech-savvy user, developer, or enthusiast, the new OpenAI models allow you to run the chatbot locally on your computer without a subscription and full control over your data.

If you’re up to the task, you can experiment, customize, and fine-tune the model since it has an Apache 2.0 license.

Usually, you would use GPT-OSS for coding help, complex problem solving, conversations, and agentic workflows. In addition, you can use it offline without an internet connection.

Who is the target audience for OpenAI GPT-OSS?

These AI models are tailored for developers, researchers, academics, and tech-savvy individuals.

If you want to run these AI models locally (independently from the cloud) on your computer, the easiest method is to use Ollama, which is an open-source tool that allows you to run Large Language Models directly on your local computer running Windows 11, 10, or another platform.

Although it’s best to install and run AI models on devices with an NPU (Neural Processing Unit), such as a Copilot+ PC with a Snapdragon processor, you can run these models on virtually any hardware. Ollama is also partnering with Nvidia to accelerate GPT-OSS on GeForce RTX and RTX Pro GPUs.

My only disclaimer is that without an NPU or capable GPU, the AI chatbot can take a very long time to answer a simple question.

In this guide, I’ll outline the steps to set up the Gemma AI models on Windows 11.

Install GPT-OSS on Windows 11 with Ollama (GUI method)
Install GPT-OSS on Windows 11 with Ollama (CLI method)

Install GPT-OSS on Windows 11 with Ollama (GUI method)

To install the GPT-OSS or 3n locally on Windows 11, follow these steps:

Open Start on Windows 11.
Search for Command Prompt (or Terminal), right-click the top result, and choose the Run as administrator option.
Type this command to install the official Ollama tool and press Enter:
```
winget install --id Ollama.Ollama
```
Click the Finish button to complete the installation.
Open the Ollama app from the Start menu if it didn’t launch automatically.
Click the AI model’s menu and select the “GPT-OSS:20b” option.
Compose and submit a question to start the download of the OpenAI GPT model.

Once you complete the steps, you can start using the GPT-OSS AI model through the command-line interface.

If you have an Ollama account, you can sign in to use the built-in web search to combine the capabilities of the model with the latest information available online.

In case you’re wondering, the Turbo option is a cloud-based service from Ollama that allows you to run large AI models using datacenter-grade hardware. It’s designed to provide high performance and speed for models that might be too large or too slow to run on a local computer. The monthly subscription is priced at $20.

The Ollama app doesn’t offer an option to pre-download the available AI models. However, you can complete this task from the command-line tool.

Install GPT-OSS on Windows 11 with Ollama (CLI method)

To install the OpenAI GPT-OSS locally on Windows 11, follow these steps:

Open Start.
Search for Command Prompt (or Terminal), right-click the top result, and choose the Run as administrator option.
Type this command to install the official Ollama tool and press Enter:
```
winget install --id Ollama.Ollama
```
Click the Finish button to complete the installation.
(Option 1) Type this command to install the smaller GPT-OSS AI model locally and press Enter:
```
ollama pull gpt-oss:20b
```
Quick note: This command downloads the gpt-oss:20b, which includes 20 billion parameters. However, you can also use the gpt-oss:latest option to download the latest model available at the time.
(Option 2) Type this command to install the larger GPT-OSS AI model locally and press Enter:
```
ollama pull gpt-oss:120b
```
(Optional) Type this command to confirm that the AI model has been installed and press Enter:
```
ollama list
```
(Optional) Type this command to show details of (for example) the smaller GPT-OSS model, such as configuration and parameters, and press Enter:
```
ollama show gpt-oss:20b
```
Type this command to run (for example) the GTP-OSS model and press Enter:
```
ollama run gpt-oss:20b
```

Once you complete the steps, you can start using the latest open-source GPT AI model from OpenAI through the command-line interface.

In addition to using the Windows Package Manager (winget) to install this tool, you can always get the Ollama installer from its official page or this GitHub page.

You can also use the ollama --help command to view other available commands, and the ollama run --help command to list the commands available for a specific model.

If you plan to run one of these AI models on your computer, you should choose the gpt-oss-20b model because of its footprint and hardware requirements.

If you want to use the gpt-oss-120b model, you’ll need very capable hardware, such as an NVIDIA GPU with at least 80 GB VRAM, server-grade processor (for example, AMD EPYC or Intel Xeon), at least 256GB of memory, and High-speed NVMe SSD with at least 500 GB free for model weights and temporary files.