Model Attribution Audit

Reverse-engineer true underlying LLM model weights through non-semantic token distribution analysis.

64 Verified Profiles
QUICK TARGETS:
Three Independent Mathematical Probes

Copy each challenge prompt into your LLM chat session (Claude, ChatGPT, etc.) and paste the unedited integer response.

Fingerprint Bank & Reference Profiles

Inspect statistical centroids, Hellinger projection matrices, and enrolled frontier models.

DeepSeek Family Models
Enroll New Model Fingerprint via API

Sample 36 challenge prompts to calibrate and store a new model's mathematical centroid.