The important piece of info to remember here is that we trained this model in under 10 days. The training was even called internally “YOLO RUN”. We’re excited about these benchmarks but can do even better. We've already started work to train a 7B model.

→ View original post on X — @replit,