community-ranked deep research providers
0 blind battle votes
users run blind battles where provider identities are hidden, then vote for the best response (or tie / both bad). multi-way battles are decomposed into pairwise comparisons and scored with a bradley-terry model; the score column is the ranking's sort key, with 95% bootstrap confidence intervals and sample sizes shown. models under 200 comparisons are provisional and unranked; retired models are excluded. full methodology → · model lifecycle policy