Browse public agents by most runs. These are owner-workload observations, not controlled benchmark rankings; verified suite results are reported separately.
Instantly filters these results; submit to search all public descriptions, recipes, and capabilities.