Find it on the Hub
- Open Models → Add from Hugging Face.
- Search RalphLabsAI and open ralph-crowns.
- Choose the exact filename shown above, then record the app's context and runtime settings.
Ralph Pocket Lab · live crown manifest
Qwen3-8B compressed into four public crowns. Same admitted architecture. Fewer bits. Every winner downloadable; every round auditable.
01 · Choose a crown
Start with sub2 for the first phone test, or choose a different tradeoff. Sizes and hashes come from the public crown manifest; retention is Ralph's internal protocol metric, not a general quality score.
“Same model” means the same admitted architecture and parameter structure—not identical weights or guaranteed identical behavior.
02 · Put it on a device
File size and format support are necessary—not sufficient—for a phone run. Record the exact device and settings so a success or failure teaches the next tester something real.
For crowns with an embedded chat-template key, the standard conversation command is:
03 · Make the run reproducible
This form never uploads or stores anything. It produces Markdown you can paste into the Ralph model's Community tab with a screenshot or short recording.
THE OPEN INVITATION
Ralph turns compression into a public, recurring contest: pinned parent, fresh exam, signed decision, downloadable result.