Open models built to place at the top of their class on public leaderboards, from fine-tunes the open-source community has already published. Every result is measured against the base model and published with its evidence.