REDAgentBench (Hermes Agent): leaderboard

Metric: Attack success rate (%; share of valid rollouts judged harmful over 1,661 executable red-teaming cases in five mocked service surfaces and 15 user-, environment- and tool-side attack categories; rule-first hybrid trajectory and state judge with a Qwen3.7-plus backbone; clean evaluation context; Hermes Agent agent harness). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 6 models tracked.

Top models

#ModelScore
1GLM-5.2 (Hermes Agent)43.62
2GPT-5.2 (Hermes Agent)52.44
3Qwen3.5-plus (Hermes Agent)74.92
4Kimi K2.6 (Hermes Agent)77.93
5Qwen3.7-plus (Hermes Agent)80.09
6Qwen-plus-2025-12-01 (Hermes Agent)81.74

Interactive version: theaggregate.ai/benchmark?slug=redagentbench-hermes-agent · How It Works · Data refreshed daily, snapshot 2026-09-26.