BrowseComp rank

Stored rank on the BrowseComp slice. 4 models store a rank, from 1 through 4 with no gaps. Rank 1 is Claude Opus 5. Rank 4 is GPT-5.6-Luna. The number is that rank. A longer bar is a higher rank number, further down the board. This is not the accuracy and not the cost per task. A model with no stored rank stays off the page.