🔧 Lesson: Pulling Models (e.g., ollama pull mistral)

🎯 Lesson Objective

By the end of this lesson, you’ll understand how to:

  • Download and manage open-source LLMs locally using Ollama

  • Pull specific models like Mistral, LLaMA, or Code LLaMA

  • Choose models based on your use case (e.g., chat, coding, summarization)

🧠 Why Pull a Model?

Before using a local LLM with Ollama, you need to download (or “pull”) it.
Pulling a model means you’re downloading a pre-trained large language model (LLM) to your machine so it can run without the internet.

Running models locally allows:

  • 💡 Full offline capability

  • 🕵️‍♂️ Private document analysis

  • 💸 No API costs (unlike OpenAI)

  • ⚡ Fast responses (if your hardware supports it)

🚀 Step-by-Step: Pulling a Model Using Ollama

✅ 1. Install Ollama

If you haven’t yet, install Ollama from https://ollama.com.

Supported OS: macOS, Windows, Linux
Installation is a one-line command (platform-specific)

✅ 2. Open Your Terminal or Command Prompt

On your machine, open:

  • macOS/Linux: Terminal

  • Windows: Command Prompt or PowerShell

✅ 3. Pull a Model

Use the command:

ollama pull mistral

This downloads the Mistral 7B model — a powerful, lightweight LLM suitable for:

  • Text generation

  • Summarization

  • Question answering

  • Document chat

  • General-purpose agent building

You’ll see output like:

✔ downloading model mistral:latest from ollama...
✔ success

✅ 4. Run the Model

Once pulled, you can run it with:

ollama run mistral

This starts a chat session with the Mistral model.

🧰 Popular Models You Can Pull

Model Use Case Command
mistral General-purpose, fast, accurate ollama pull mistral
llama3 Meta’s LLaMA 3 model ollama pull llama3
gemma Google’s lightweight model ollama pull gemma
codellama Code-focused tasks ollama pull codellama
deepseek-coder Strong code generation ollama pull deepseek-coder
orca-mini Small + fast (low RAM) ollama pull orca-mini

⚠️ Model Size & RAM

Model Size RAM Required (approx.)
Mistral ~4–8 GB 8–16 GB
LLaMA 3 ~8–12 GB 16+ GB
Code LLaMA 8–12 GB 16+ GB
Orca Mini 3–5 GB 4–8 GB

If your system has low RAM, prefer smaller models like:

ollama pull orca-mini

🧪 Troubleshooting

Issue Solution
“Ollama not found” Reinstall Ollama and ensure it’s added to your PATH
“Model pull stuck” Check your internet or try a different mirror
“High RAM usage” Try a smaller model like orca-mini or use quantized versions

🎓 Exercise

  1. Open your terminal

  2. Run: ollama pull mistral

  3. Then: ollama run mistral

  4. Ask the model: “Explain how a car engine works in simple terms.”

✅ Summary

  • Use ollama pull model-name to download models

  • Run them with ollama run model-name

  • Choose models based on performance and RAM

  • Pulling models is essential before using them in your AI agents

  •  

61
PreviousNext