# DeepSeek Deployment

***Copyright © Quectel Wireless Solutions Co., Ltd. 2026. All rights reserved.***

---

This document demonstrates how to deploy DeepSeek locally on the **Quectel Pi M1** smart mainboard.

**DeepSeek** is a high-performance open-source large language model developed by a Chinese team. It offers strong capabilities in natural language understanding, code generation, and logical reasoning, runs efficiently in local environments, and is well suited for use cases such as intelligent Q&A, text generation, and code assistance.

## **Prerequisites**

1. The Debian system on the **Quectel Pi M1** has been successfully flashed and booted;

2. The system can access the external network.

## **Update Software Sources**

Run the following command in the terminal:

```plaintext
sudo apt update
sudo apt upgrade
```

> Before installing software, it is recommended to update the system packages to ensure dependencies are at the latest versions.

```{image} images/image_Hkmkbvkdooem4Dx1QIWcrOZHnHe.webp
:width: 557px
:height: 90px
```

## **Install Basic Tools**

```plaintext
sudo apt install curl
```

> `curl` is a common command-line download tool that will be used when installing Ollama later.
> If the system indicates that `curl` is already installed, you can skip this step directly.

```{image} images/image_C8ORbymIHokGe8xqaJIc4665nKb.webp
:width: 526px
:height: 81px
```

## **Install Ollama**

Ollama is a lightweight model runtime environment that can deploy and run various open-source models locally.

Run the following command in the terminal:

```plaintext
curl -fsSL https://ollama.com/install.sh | sh
```

```{image} images/image_JPx3b9dmfovivWxlDamc1l6Mncc.webp
:width: 885px
:height: 255px
```

## **Start the Ollama Service**

Ollama requires a server process running in the background to manage models.

Open a terminal and execute the following command:

```plaintext
export OLLAMA_HOST=0.0.0.0:11435
ollama serve
```

> This terminal window must remain open.

## **Run the DeepSeek Model**

Open **another terminal window** and run the following command in it to pull and start the **DeepSeek-R1 1.5B** model:

```plaintext
ollama run deepseek-r1:1.5b
```

```{image} images/image_OzcSbqoGsoHEH6xPmg0c3KRunfd.webp
:width: 658px
:height: 223px
```

You can enter text in the red box area to start a conversation, for example:

```{image} images/image_V4Ebbz578oe4Vcxw4PPc9qb1n9g.webp
:width: 444px
:height: 82px
```

## FAQ

Download error: ERROR: This version requires zstd for extraction. Please install zstd and try again:

```plaintext
# Manually install the zstd library
apt update 
apt install zstd
```

After running ollama serve, you see: bash: ollama: command not found

```bash
# Clean up old remnants first
rm -rf /lib/ollama 
rm -f /bin/ollama /usr/bin/ollama 
rm -f /etc/systemd/system/ollama.service

# Download the package to disk first, do not pipe
curl -L https://ollama.com/download/ollama-linux-arm64.tar.zst -o /tmp/ollama.tar.zst

# Manually extract
tar -I zstd -x -f /tmp/ollama.tar.zst -C /usr

ls -la /usr/bin/ollama 

ollama --version
```