The harness is the scaffold. Compare them on real numbers.

SWE-bench Verified resolution rate × GitHub repo health × release cadence. Refreshed every 15 minutes from public APIs. No accounts, no tracking, no fake data.

—harnesses

—SWE-bench submissions

—events tracked

—last refresh

#	Harness	SWE-bench Verified	★ stars (Δ7d)	commits 30d	open issues	latest release	grade
Loading leaderboard…

Diff feed

Reverse-chronological log of new SWE-bench submissions, releases, and star milestones across the arena.

Loading…

Compare two harnesses

vs

Methodology

Every column on the leaderboard maps to a real, public, runtime-fetched endpoint. Zero seeded scores. If a value has not been fetched yet, the cell shows an em-dash.

Loading…