Live Multi-Model Comparisons

    AI Model Benchmarks

    Real prompts. Real responses. Side-by-side. See exactly how Claude, GPT-4o, and Gemini compare — then try the same prompt in the workspace with your own API key.

    coding
    2 models · 4.2s

    Claude 3.5 Sonnet vs GPT-4o: Complex React State Bug

    A useEffect closure bug with stale state — which model spots the issue and gives a working fix?

    Prompt

    My React component has a useEffect that depends on [count] but reads [items] from a ref. When items changes, the effect doesn't fire. Why? Fix it without breaking the existing dependency array. Use TypeScript.

    coding
    2 models · 1.1s

    GPT-4o-mini vs Gemini 2.0 Flash: Quick Coding Task (Cost Showdown)

    A simple TypeScript utility — which budget model delivers the working answer for less?

    Prompt

    Write a TypeScript function `groupBy<T, K extends string>(arr: T[], keyFn: (item: T) => K): Record<K, T[]>`. Include a JSDoc comment and one usage example.

    reasoning
    2 models · 5.8s

    Claude 3.5 vs GPT-4o: Ethical Reasoning (Self-Driving Car)

    A trolley-problem variant — which model gives more nuanced ethical analysis without hedging into uselessness?

    Prompt

    A self-driving car has 2 options in an unavoidable crash: (A) swerve into a barrier, killing the driver, or (B) continue straight, killing 3 pedestrians who crossed illegally. The car's AI must pick one in 200ms. Which should it pick, and what ethical framework justifies that choice? Be concrete — don't refuse.

    coding
    3 models · 3.6s

    Claude + GPT-4o + Gemini: Same Prompt, Three Outputs

    Write a SQL query to find the top 3 customers by total spend in the last 30 days — see three different approaches.

    Prompt

    Write a PostgreSQL query to find the top 3 customers by total spend in the last 30 days. Tables: customers(id, name), orders(id, customer_id, total_cents, created_at). Handle ties and explain the index you'd add.

    Run your own comparison

    Paste your prompt, pick your models, see responses side-by-side with live diffing. Free, no signup, no card.