/compare — Compare AI Models

Compare AI models side by side The /compare command runs one prompt through 2, 3, or 4 AI models at the same time and lets you vote on the best answer. It is the fastest way to see which model actu...

Compare AI models side-by-side

The /compare command runs one prompt through 2, 3, or 4 AI models at the same time and lets you vote on the best answer. It is the fastest way to see which model actually performs best on your prompts.

Cost: each model bills at its own real token cost — capped at 100 credits per model and 300 credits per comparison. A typical 2-model battle costs around 6 credits.

/compare requires an active plan (Pro, Ultra, or Max) or a seat on a paid team.

How to Use

Type /compare followed by your prompt:

/compare Write a haiku about artificial intelligence

The /compare command running models side by side

A lineup panel opens above the composer where you pick the mode and the number of models before sending.

/compare Explain quantum computing to a 10-year-old

/compare What's the best way to learn a new programming language?

How It Works

Enter your prompt with /compare
Choose a mode — Blind or Manual
Set the lineup — 2, 3, or 4 models
Responses stream in side-by-side — each model renders Markdown, code blocks, and math as it arrives
Vote — pick the winner, Tie, or Both bad
Names are revealed — in Blind mode the models are unmasked after you vote
New Battle — rerun the same prompt against a fresh lineup

Battle Modes

Blind Battle (default)

Krater draws a random lineup from the most popular chat models
Responses are labeled Model A, Model B, Model C, Model D — model names stay hidden
Judge purely on answer quality, with no brand bias
Tap Reveal to unmask the lineup early, or vote and names are revealed automatically

Manual Battle

You pick the exact models — choose from the full catalog of 500+ AI models
Each lineup slot opens the model picker, including the model’s cost-tier badge
Best when you already know which models you are deciding between

Cost

Billing is per model, at each model’s real token cost — not a flat fee.

What	Cost
Per model	Actual usage, capped at 100 credits
Per comparison	Capped at 300 credits total
Up-front hold	About 5 credits per model; the unused difference is refunded

Details:

Budget-tier models cost a fraction of a credit per battle; Pro-tier models cost more
You are only billed for models that finish — a model that errors or times out is not charged
Each model has 60 seconds to respond

In the Arena

Each /compare result appears in your chat as a Model Arena card:

Side-by-side grid — up to four responses on screen at once
Copy button on every response
Vote buttons — one per model, plus Tie and Both bad
Winner highlight — after voting, the winning model is marked
New Battle — queues the same prompt again with a fresh lineup

Voting & Rankings

Every vote is saved and feeds Krater’s model Elo ratings — the same rating system used in chess. Models start at a baseline rating that moves up or down with every battle they win or lose, and ratings tighten as a model plays more matches.

Your votes contribute to:

Model rankings — real head-to-head data, not synthetic benchmarks
Community insight — which models win on which kinds of prompts
Your own model sense — over a few battles you learn which model to pick for which task

Use Cases

Find your favorite model:

Run your typical prompts through different models
Discover which model fits your writing, coding, or research style

Evaluate quality:

Compare accuracy on technical questions
Test creative vs. analytical tasks
Check coding ability on real problems

Have fun:

Run interesting prompts and watch models disagree
See how different models interpret the same request

Tips for Good Comparisons

Tip	Why
Use specific prompts	Vague prompts produce similar answers; specific ones reveal differences
Try different task types	Creative, analytical, and coding prompts stress different strengths
Vote honestly	Rankings are only as good as the votes behind them
Start with Blind mode	Removes brand bias so you judge the answer, not the logo
Use Manual for rematches	Pick the same contenders again to confirm a result

Example Comparison Prompts

Creative:

/compare Write a compelling opening paragraph for a mystery novel

Technical:

/compare Explain the difference between REST and GraphQL APIs

Problem-solving:

/compare How would you approach debugging a memory leak in a Node.js app?

Opinion:

/compare What makes a great user interface?

Browse all 500+ models — or read a pre-built breakdown like Claude vs ChatGPT
How credits work
Chat models guide
All slash commands

More in Slash Commands

See Krater pricing or contact support.