Agent Lore

Why model choice matters

Faster models cost less and respond quicker. Powerful models handle complex reasoning better. Picking the right one saves money and time.

Model tiers

In Claude Code

# Start with a specific model
claude --model claude-sonnet-4-20250514

# Inside a session, switch with /model
/model claude-haiku-3-5-20241022

In Cursor

Open the model selector in the chat panel (bottom-left dropdown). Each conversation can use a different model.

Rules of thumb

• Use fast models for: autocomplete, simple questions, formatting, small edits
• Use standard models for: feature implementation, code generation, refactoring
• Use powerful models for: debugging complex issues, architecture planning, multi-step reasoning

Cost impact

A fast model can be 10-20x cheaper per token than a powerful one. Using Haiku for simple tasks instead of Opus across a full day can cut costs significantly.