Model Guide

Llama 3.2 Models

Text Models

llama3.2:3b (Default - Recommended for Workshop)

• Size: 2.0 GB download
• Parameters: 3 billion
• Context: 128K tokens
• RAM: 4-6 GB required
• Use Cases: General purpose, instruction following, summarization
• Download: ollama pull llama3.2

llama3.2:1b (Faster Alternative)

• Size: 1.3 GB download
• Parameters: 1 billion
• Context: 128K tokens
• RAM: 2-3 GB required
• Use Cases: Quick responses, resource-constrained systems
• Download: ollama pull llama3.2:1b

Vision Models (Advanced)

llama3.2-vision:11b

• Size: 7.9 GB download
• Capabilities: OCR, image analysis, visual Q&A
• RAM: 8+ GB required
• Download: ollama pull llama3.2-vision

Embedding Models

nomic-embed-text (Recommended)

• Dimensions: 1,024
• Context: 8,192 tokens
• Size: 274 MB
• Use: Semantic search, RAG
• Download: ollama pull nomic-embed-text

mxbai-embed-large (Alternative)

• Dimensions: 1,024
• Size: 669 MB
• Use: High-quality embeddings
• Download: ollama pull mxbai-embed-large

Other Useful Models

codellama:7b (Code Generation)

ollama pull codellama:7b

mistral:7b (Alternative LLM)

ollama pull mistral:7b

llama3.1:8b (Previous Version)

ollama pull llama3.1:8b

Choosing the Right Model

Model Management

List installed models:

ollama list

Remove a model:

ollama rm llama3.2:1b

Update a model:

ollama pull llama3.2  # Downloads latest version

Check model details:

ollama show llama3.2

Quantization Levels

Models are available in different quantizations (size vs. quality trade-offs):

• Q4_K_M: Default, balanced (recommended)
• Q5_K_M: Larger, better quality
• Q2_K: Smallest, lower quality

Example:

ollama pull llama3.2:3b-q5_k_m  # Higher quality quantization

See Quantization for more details.