Feature: Automatic Model Selection for Token Efficiency

Problem

Agents default to expensive models (Opus/Sonnet) for tasks that can be completed with cheaper models (Haiku). This wastes user tokens unnecessarily.

Request

Implement automatic model selection that always uses the minimum capable model for the task:

• Haiku: Exploration, search, file reading, research
• Sonnet: Context-gathering, moderate analysis
• Opus: Only when complex reasoning is required

Benefit

Token efficiency should be the default behavior. Users shouldn't need to manually optimize—always pick the cheapest model that works for the given task.