WhatsMyLLM 中文

What LLM am I using?

Send the prompt below. Paste the reply to compare models.

Free · Browser-local · No sign-up

Test your model

Your replies · edit or add

Use the same model in a fresh chat for each prompt. No code or tools.

Prompt 1 · 301 numbers

Prompt 1

301 numbers
View prompt
For each of 301 positions, make one separate first-instinct choice of an integer from 1 to 355 inclusive. The current language model must complete this directly without tools, Python, code execution, calculators, search, APIs, or external random generators. Choose every position separately. Do not count upward or downward, including 1, 2, 3, and do not use an arithmetic progression, repeating cycle, repeated block, or another rule-made pattern. Accidental repetitions are valid. Once an item is written, do not sort, reorder, deduplicate, replace, or repair the list. Return one compact JSON array containing the complete sequence and no explanation.

0 / 3 usable
Options & input help

Prompt format

Prompt format

This changes the prompt, not the paste format. Replies are kept separately for each prompt format.

Optional. Adds an explanation without changing the fingerprint result.

Paste the reply itself, a number array, or one completed Chat Completions / Messages response JSON. Response text is extracted automatically; no key or endpoint is needed.

Ctrl / ⌘ + Enter to compare

Your result

Result preview

  • Most likely model
  • Fit & separation
  • All 14 candidates

Start with one reply. Add two more for a fuller comparison.

A statistical match, not identity verification. Method & limitations

14 models in the bank

Find yours or inspect its evidence

Know what this test can establish

It compares replies to these exact prompts—not essays or a model’s self-description. In the measured paste flow, 38 of 42 runs named the right model, 4 stopped at the family, and 0 named a wrong model. These are test results, not a guarantee for your service.

Unlisted versions can resemble listed siblings. With Opus 4.7 removed, 7 of 12 three-reply tests named its nearest sibling instead. The test cannot identify a provider, channel or subscription. Inspect the full evidence.

Open full page