The A.I. Daily
Harness leaderboard benchmarks Kimi Code, Claude Code on terminal tasks
A Reddit community post introduced a Harness leaderboard benchmarking coding agents including Kimi Code, Hermes, Pi, Claude Code and Codex against Terminal Bench 2.1 and Deep SWE 1.1. It is a community-run comparison.