Compare Mode

Run prompts on multiple models and compare responses

Compare mode sends the same prompt to multiple AI models simultaneously and displays their responses side-by-side for easy comparison.

Usage

TUI (Interactive)

/compare <agents> "<prompt>"
/sequential      # toggle sequential execution
/pick            # toggle LLM picks best

CLI

puzldai compare "<prompt>" -a <agents> [-s] [-p]

Flag	Description
-a	Comma-separated agents to compare
-s	Run sequentially instead of parallel
-p	Have an LLM pick the best response

Examples

# TUI: Compare Claude and Gemini
/compare claude,gemini "Explain microservices architecture"

# CLI: Compare three models with pick mode
puzldai compare "Write a binary search function" -a claude,gemini,ollama -p

Features

Side-by-Side View

Responses are displayed in columns, allowing you to easily compare:

• Response quality and completeness
• Different approaches to the same problem
• Response time for each model

Sequential Mode

By default, all models run in parallel. Enable sequential mode to run one at a time:

/sequential    # Toggle sequential mode

This is useful when you want to see responses as they complete.

Pick Mode

Enable pick mode to have an AI select the best response:

/pick          # Toggle pick mode

When enabled, after all responses are collected, an AI evaluates them and selects the best one with an explanation.

Project Context

Compare mode automatically injects project structure context, so models understand your codebase without needing file access.

Keyboard Shortcuts

Key	Action
Escape	Hide view (keeps running in background)
Ctrl+E	Return to compare view
Ctrl+C	Cancel all requests
←/→	Navigate between responses

History

After exiting compare mode, results are saved as a compact view in chat history. Press Ctrl+E to expand and view again.

Options

Option	Command	Description
Sequential	/sequential	Run models one at a time
Pick	/pick	Auto-select best response