← Back to Blogs

Run Large Language Models Locally from Scratch

Run Large Language Models Locally from Scratch

How to Run Large Language Models Locally from Scratch?


Running Large Language Models (LLMs) locally is becoming more accessible — giving you privacy, speed, and control over your AI workflows without relying on cloud APIs.


In this guide, we'll set up everything from scratch using:

  • Conda for environment management
  • Pip for installing Python dependencies
  • Ollama for running local LLMs like Mistral-7B and DeepSeek

1. Why Run LLMs Locally?

  • Privacy – Keep sensitive data on your machine.
  • Offline Capability – No internet needed after setup.
  • Cost Savings – Avoid API subscription costs.
  • Customizability – Fine-tune and experiment without restrictions.

2. Install Conda

If you don’t have Conda yet, download Miniconda (lightweight) or Anaconda.

After installation, verify:

conda --version

3. Create a virtual environment

We'll be creating a virtual environment for our LLM project.

conda create --name local-llm
conda activate local-llm

4. Install Ollama

Ollama is a runtime for local models which is available on macOS, Linux, and Windows.

In macOS:

brew install ollama

In Linux:

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

In Windows:
Download from Ollama's [official website] (https://ollama.com/download)

5. Verify Ollama Installation

ollama --version

If installed correctly, you should see the Ollama version number.

6. Download Mistral-7B and DeepSeek Models

Now we'll be downloading two models locally. Ollama makes it simple to pull models:

ollama pull mistral
ollama pull deepseek

7. Useful Ollama commands

Here are some list of useful Ollama commands:

ollama list # list all available models
ollama run <model_name> # run any specific model interactively
ollama serve # serve ollama api so that other apps can connect
ollama rm <model_name> # remove model from your system

8. Integrating Ollama with Python

Install the python client:

pip install ollama

Example script:

import ollama
response = ollama.chat(
    model="mistral",
    messages=[{"role": "user", "content": "Write a one line description of Australia"}],
)
print(response["message"]["content"])

Output from the model:

Australia, the world's largest island and sixth-largest country, is a vibrant and
diverse nation known for its unique wildlife, vast outback landscapes, bustling cities,
and beautiful beaches.

Here's a fun fact: Did you know that Australia has more types of terrestrial mammals
than any other continent? This includes the famous kangaroo, wallaby, koala, and Tasmanian
devil!

Happy Coding 💻

Understanding Software Design Patterns...