Pulling and Running Models

The Model Library

Models live in the Ollama library under names with tag suffixes. Common families include llama3.2, qwen2.5, mistral and phi3.

Pull a Model

ollama pull llama3.2

Downloading separately lets you give a big model time to finish before using it. Running already pulls if needed.

Tags and Sizes

• llama3.2:1b - Tiny, fast, low quality.
• llama3.2 - 3B, a solid everyday default.
• qwen2.5:7b - Strong mid-size general model.
• :q4 / :q8 - Quantization levels of the same weights.

Larger models answer better but need more memory and run slower.

Run Interactively

ollama run qwen2.5:7b

Type messages at the prompt like any chat, or use /exit to leave. The REPL keeps a local conversation.

Key Points

• ollama pull downloads a specific tag.
• ollama run starts a chat and pulls on demand.
• Tags encode size and quantization.
• Pick a tag that fits your memory before searching for quality.