🤖 Ollama Beginner Guide: Running & Controlling Local LLMs
In this tutorial, we will learn how to use Ollama to run Large Language Models (LLMs) locally on your system. This guide is beginner-friendly and covers everything from running models to customizing their behavior.
🚀 What is Ollama?
Ollama is a tool that allows you to run AI models like LLaMA, Qwen, and Gemma locally on your machine without needing cloud APIs.
- ✔ No internet required after setup
- ✔ Full control over models
- ✔ Faster and private execution
⚙️ Initializing Ollama
Before using any model, you need to download and run it.
ollama pull llama # Download the model
ollama run llama # Run the model
If the model is not already downloaded, ollama run will automatically download it.
💬 Talking to the Model
Once the model is running, you can interact with it like a chatbot.
/?→ Show help commands/clear→ Clear chat history/bye→ Exit the model
⚡ Running One-Time Queries
If you don’t want to maintain conversation history, you can run a single query directly:
ollama run llama "Enter your query"
This is useful for quick tasks like summarization or Q&A.
🖼️ Working with Image Models
Ollama also supports image-based models for tasks like summarization.
ollama run gemma3:4b
Then provide an instruction like:
Summarize "C:\path\to\your\image.jpg"
This allows the model to extract and explain information from images.
🧠 Thinking Models
Some models like Qwen or DeepSeek-style models have thinking capabilities.
- ✔ Show reasoning steps
- ✔ Better for coding and problem solving
- ✔ More accurate for complex tasks
🔍 Model Exploration Commands
1️⃣ /show Command
Used to explore model details and capabilities:
/show
Available options include:
/show info
/show license
/show parameters
/show system
/show template
2️⃣ /set Command
Used to customize how the model behaves.
/set system "You are a funny assistant"
This sets a system instruction that affects all responses.
📌 Example Output
User: hi
Model: Hey there! Ready for some laughs? 😄
🧾 JSON Mode (Very Important)
You can force the model to return structured output using JSON format:
/set format json
Example:
{
"name": "Atul Kushwaha",
"institution": "KIET College, Ghaziabad",
"year": 1,
"stream": "AIML"
}
This is very useful for APIs and automation.
💾 Saving Your Own Model
You can customize a model and save it for future use.
Steps:
/set parameter temperature 0.5
/save MyModel
Now your model is saved and can be reused without reconfiguration.
📌 View Models
ollama ls
❗ Important Note
Saved models may include previous conversation history. This can affect future responses.
To avoid this, use:
/load MyModel
🗑️ Deleting a Model
ollama rm MyModel:latest
This removes the model from your system.
⚠️ Limitations
- ✔ Large models require high RAM
- ✔ Initial download size can be large
- ✔ Saved models may retain history
🚀 Conclusion
Ollama is a powerful tool for running AI models locally. It gives you full control over how models behave and allows you to build AI-powered applications without relying on external APIs.