swe-pulse

live SWE-bench cross-variant tracker
# Submission Org Resolved % $/instance Attempts Date OSS

Each row is a normalized model family that appears on at least two leaderboards. Cells show the best resolved score that family achieved on each variant. Empty cells mean no submission exists — they are not interpolated.

X = instance cost (USD, log scale). Y = resolved %. Filled = Pareto-optimal.

Submissions per month across all leaderboards (stacked by variant). Pulled from the date field of each entry in the upstream leaderboards.json.