Creating Custom AI Models with Modelfiles

Post 1 hours ago - 08 Jun 2026 | Updated 08 Jun 2026 | 10

Introduction

In the previous blog, we learned what Modelfiles are and how they help customize the behavior of AI models. However, understanding Modelfiles is only the first step. The real use of Modelfiles is when we ued them to create our own custom AI models.

By using Modelfiles we can convert a general purpose AI model into a AI assistant to perform any specific task. For example we can create AI Teacher, Coding Assistant, Interview Coach, Customer Support Agent, or a personal AI assistant without retraining the original model.

Why Create Custom AI Models?

Normally when we run a model such as Llama3 it behaves like a general purpose AI assistant.

Workflow:

user → llama3 → general response

The model tries to answer every type of question without following any specific role. However in real world applications we need specialized AI assistants.

Examples:

AI Teacher
Coding Assistant
Interview Coach
Customer Support Assistant
Personal Tutor
Research Assistant

Instead of writing the same instructions repeatedly in every prompt we can create a custom model that permanently follows those instructions. This is where Modelfiles become useful.

Step 1: Create a Modelfile

The first step is creating a Modelfile.

Example:

FROM llama3

SYSTEM You are an AI teacher who explains every concept in short.

Understanding the Modelfile

FROM Instruction

The FROM instruction specifies the base model that will be used.

Example:

FROM llama3

This means our custom model will be built using Llama 3.

SYSTEM Instruction

The SYSTEM instruction defines the personality and behavior of the model.

Example:

SYSTEM You are an AI teacher who explains every concept in short.

Now the model will try to answer like a teacher and provide short explanations.

Step 2: Save the Modelfile

Create a file named:

Modelfile

and save the content inside it.

Example:

FROM llama3

SYSTEM You are an AI teacher who explains every concept in short.

Step 3: Create the Custom Model

After creating the Modelfile use the following command:

ollama create myteacher -f Modelfile

Understanding the Command

ollama create :- Used to create a new custom model.

myteacher:- This is the name of the new model.

-f :- Specifies the Modelfile location.

Workflow:

Modelfile → ollama create → custom model

Ollama reads the instructions from the Modelfile and creates a new customized model.

Step 4: Run Custom Model

Once the model is created then it can be run using:

ollama run myteacher

Now your custom AI model is ready to use.

Testing Model

let any question to ask like:

What is Artificial Intelligence?

Possible response:

Artificial Intelligence (AI) is a field of computer science that enables machines to simulate human intelligence and perform tasks such as learning, reasoning, problem-solving, decision-making, and understanding language.

Notice that the response style follows the teacher behavior defined in the SYSTEM instruction.