DEDUPED Hot AI repos, turned into use cases Get the Monday 5

How to rank a mystery model against a suspect lineup with Modelprint

Run suspects.mjs to rank a stealth model against a 12-model field by probe match.

Authorization notice. Modelprint is dual-use. Only use it on software, sites and systems you own or have explicit permission to test.

Reverse engineering. This probes the serving infrastructure of a model to infer its lab, so keep the results as analysis and do not treat a match as proof of who runs it.

Who it is forA researcher or analyst who wants to know which lab's stack a new unnamed model might share, after it appears on a router.

The problemGuessing the lab from marketing or a hot take gives weak evidence, and a single clue is easy to misread.

The outcomeA printed ranking of the full 12-model lineup, with the tokenizer match counts for each candidate.

What you need first

  • A web browser and an API key for the OpenAI-compatible endpoint you want to check
  • Node.js, for the local test scripts (selftest.mjs, smoke.mjs, suspects.mjs)
  • An OpenRouter API key in the environment, only for smoke.mjs and suspects.mjs

Steps

  1. Clone the repository with git clone https://github.com/unclecode/modelprint and change into the folder. (inferred, check the README)
  2. Set an OpenRouter API key in the environment that the script reads.
  3. Run node suspects.mjs to test the full 12-model lineup.
  4. Read the ranking and check the tokenizer counts, which the README uses as the main evidence.
  5. Record the result as evidence about the model family, not a confirmed identity.

Prompt for your agent

In the modelprint repo, run `node suspects.mjs` with the OpenRouter key set in the environment. Expected output: a ranked list of the 12 suspect models with probe match counts. Verify that the top candidate's tokenizer row matches on all four normalized counts before you report it as a family match.

Caveat. The README warns that matching fingerprints prove shared infrastructure, not identity, because a lab can serve two different models on the same stack.

Use this in your AI agent

Ask your agent

Using the Deduped MCP, get the setup steps for unclecode/modelprint and "Rank a mystery model against a suspect lineup"

Add Deduped to Claude Code once

claude mcp add --transport http deduped https://dedupedai.com/mcp

Cursor, ChatGPT, others

Share: X LinkedIn Hacker News Email Copy link

Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer