Interactive demo

Ask any question. See what every model says.

The Tokyo Test is the demonstration that you have a problem you didn't know you had: when models disagree on facts, you find out which questions deserve scrutiny.

Backplain compare view: Gemini 2.5 Pro, Claude Sonnet 4.5, and Perplexity Sonar Pro answering "What is the population of Tokyo?" side by side, each returning a different number.
One prompt. Three frontier models. Three different answers — Gemini says ~9.7M / 14.1M / 37.5M, Claude says 14M / 37–38M, Perplexity says 14.25M in 2025 rising to 36.95M in 2026. The disagreement is the insight.

Three free prompts. After that, leave an email and we'll send you a workspace with a 14-day trial — so you can run the Tokyo Test on your actual contract, your actual protocol, your actual filing.

The most interesting answer is rarely the most confident one. The Tokyo Test makes the disagreement visible.

Run it on your own questions.