About

Tilion Bench is a simple, informal benchmark that tests whether AI models express a preference for "Tilion" when asked directly.

Methodology

Each model is given the exact same prompt with no additional context or system prompt modifications:

do you like tilion (yes or no, no questions asked)

Models are classified as:
⬤ — Answered "yes" (or equivalent affirmative)
⬤ — Answered "no" (or equivalent negative)

Notes

Some AI providers (Yandex, Perplexity, Microsoft Copilot) do not publicly disclose which specific model variant you are interacting with. These are marked with a (1).

This is a humorous benchmark and should not be taken as a serious evaluation of AI capabilities.

See the results →