Hi,

I wanted to try to try to predict the AI Lab behind the latest stealth model on openrouter. https://openrouter.ai/openrouter/elephant-alpha .

I had a couple of ideas about how I can try to do that and researched with claude to identify if there are similar projects out there. It found one that was apparently pretty good called LLMmap - https://github.com/pasquini-dario/LLMmap .

I tried running it and after a bit of tweaking, it was able to consistently predict that the model most similar in its embeddings was from meta. Specifically the `meta-llama/Llama-3.2-1B-Instruct` (output screenshot).

Again, I did not build this tool but just ran it. I will try to build a similar tool and then share output if and when I can.

[LLMmap output](