Hermes Index Hermes Index is a new leaderboard from Nous Research that averages scores across HermesBench and three other agent benchmarks (Terminal-Bench 4.0, Terminal-Bench-Science, and SkillsBench) to rank AI models by both performance and cost-per-task inside the Hermes Agent environment. It uses a consistent harness with high reasoning where available, helping users pick the best model for a given time or budget and showing labs how their models actually perform in agentic workflows.