Model Management & Organization

Overview

Effective model management keeps your Ollama installation fast, organized, and storage-efficient. Learn to pull the right models, create custom variants, manage disk space, and organize models for different development tasks.

Why This Matters

• -Storage control — models range from 2GB to 40GB each
• -Fast switching — organized models let you switch contexts quickly
• -Team alignment — consistent model versions across developers
• -Performance — right model for the right task

How It Works

Step 1: Pull and List Models

# Pull a model from the Ollama registry
ollama pull llama3.1:8b-instruct-q5_K_M

# Pull a coding-specific model
ollama pull qwen2.5-coder:7b

# List all downloaded models
ollama list

# Show model details
ollama show llama3.1:8b-instruct-q5_K_M

Step 2: Create Custom Variants

# Create a Modelfile for your custom variant
cat << 'EOF' > Modelfile.typescript
FROM qwen2.5-coder:7b
SYSTEM "You are a TypeScript expert. Generate strict, type-safe code with proper error handling."
PARAMETER temperature 0.2
PARAMETER num_ctx 8192
EOF

# Build the custom model
ollama create ts-coder -f Modelfile.typescript

# Test it
ollama run ts-coder "Write a function to debounce API calls"

Step 3: Copy and Version Models

# Copy a model (useful for creating variants)
ollama cp qwen2.5-coder:7b my-coder:v1

# After tuning, create a new version
ollama create my-coder:v2 -f Modelfile.v2

Step 4: Manage Storage

# Check model storage location
# Linux/Mac: ~/.ollama/models
# Windows: C:\Users\<user>\.ollama\models

# Remove unused models
ollama rm codellama:7b
ollama rm llama2:13b

# Check disk usage by model
ollama list  # Shows size column

Best Practices

• -Pull quantized models (Q4_K_M, Q5_K_M) — best speed/quality ratio
• -Name custom models descriptively: ts-coder, python-reviewer, sql-helper
• -Remove models you haven't used in 30+ days
• -Use instruct/chat variants for interactive coding, base models for completion
• -Pin model versions in team documentation to ensure consistency

Common Mistakes

• -Pulling multiple sizes of the same model (wasteful)
• -Using the :latest tag (breaks reproducibility when model updates)
• -Not removing old models (disk fills up silently)
• -Using base models for chat tasks (poor instruction following)