One prompt. One shot. Different AI models.

• Solar System – An interactive 3D Solar System with real orbital data, time controls and a camera that flies to the planets.
• GTA-style game – A playable open-world city game in the style of the first GTA: cars, pedestrians, police with wanted stars, missions and a minimap.
• Monitor test – A professional monitor test: dead pixels, gradients, black and white level, gamma, sharpness and backlight uniformity.
• New York & Central Park map – An interactive illustrated map of Manhattan with a detailed Central Park, 30+ landmarks, search and a guided tour.
• Chess with an engine – Complete chess with every rule and its own chess engine; the move generator is verified with an objective perft test.

• 106 of 129 outputs worked on the first attempt (82 %).
• The chess move generator passed every perft test for 13 of 23 models.
• One output cost between $0 and $4.33 at API list prices.
• GPT-6 Astra and GPT-6.1 Sol have the highest average technical quality (99 of 100).
• Mistral Medium 3.5 has the best automatically measured UX (93 of 100).
• Mistral Large 4 (4/5) does not have every prompt yet – more outputs will follow.

Summary by model as of 17 October 2026. Technical quality is an automated 0–100 measurement (Lighthouse, errors during use, smoothness, mobile). UX is an automated 0–100 measurement (button size, layout on phone, tablet and desktop, keyboard use, click feedback, text readability).

See the detailed model leaderboard, output comparison and the blind arena in the menu.

The prompt is in English, identical for every model, and carries the same technical rules: one HTML file, no external libraries or images.

The model gets the prompt in a fresh conversation with no history – Claude models in Claude Code, OpenAI models in Codex, DeepSeek and Gemini directly through the API – at medium reasoning effort (or the default where a model has no medium level). Its first answer is saved without any edits, and we record time, token count and cost at the dated API list price.

Function and quality are measured objectively: running in the browser, errors during use, Lighthouse, an automated UX measurement at three screen sizes and the perft chess-rules test. Visitors vote 👍 / 👎 and pick the better of two outputs in the arena.