Tip: Cursor 3's /best-of-n — Run Same Task Across Multiple Models in Parallel Worktrees Then Compare

Cursor·high signal

Cursor 3's /best-of-n command runs the same task in parallel across multiple models, each in its own isolated git worktree, then displays outcomes side by side for comparison. This enables direct head-to-head evaluation of different models on your actual codebase rather than relying on synthetic benchmarks. Combine with Agent Tabs to view all parallel runs simultaneously in a grid layout, and /worktree for isolated single-model experiments.