Raw AI Lab · local-model research, running on this Mac

connecting…

Right now

Waiting for the lab…

This training run

No run in progress.

Every experiment so far

Validation score in bits per answer byte. Lower is better. Click a point for the full story.

Where the best model stands

Bits per answer byte by task, baseline against the current best. Shorter bars are better.
BaselineCurrent best

Experiments

Idea queue

Lab log