Skip to content

Hermes Index: agentic capability and cost per task inside Hermes Agent

An index of agentic capability from Nous Research. It averages pass@1 scores on four suites, Hermes Bench, TerminalBench 4, TerminalBench Science and SkillsBench, with every run inside the Hermes Agent harness, and reports the mean cost per task beside the score.

16 models

Higher is better.

  1. 1Claude Opus 5.5Anthropic63.3
  2. 2GPT-6.1 SolOpenAI57.1
  3. 3GPT-6 AstraOpenAI56.3
  4. 4Claude Sonnet 5.5Anthropic53.1
  5. 5GPT-6 SolOpenAI44.1
  6. 6Grok 4.7xAI39.3
  7. 7GLM-5.3Z.AI39.2
  8. 8Kimi K3Moonshot AI39.2
  9. 9Gemini 3.8 FlashGoogle38.1
  10. 10Claude Haiku 5.5Anthropic37.0
  11. 11Qwen3.8 MaxAlibaba36.2
  12. 12GLM-5.3-FlashZ.AI35.0
  13. 13GPT-6 LunaOpenAI33.9
  14. 14Step 5 PreviewStepFun32.8
  15. 15Hy4 previewTencent26.1
  16. 16Ling 3.0 FlashInclusionAI21.6
Higher is better. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric
Mean pass@1 score across four suites, with mean cost per task
Tasks
Four suites: Hermes Bench (150 tasks), TerminalBench 4 without its 4 GPU tasks, TerminalBench Science and SkillsBench
Authors
Nous Research (2026)

Scores are as reported by model developers or evaluators and may use different settings (for example reasoning effort or tools). Compare with care.