Hermes Index: agentic capability and cost per task inside Hermes Agent (Hermes Index)
We show this table for reference; we do not rank on it.
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.
Hermes Index on Hermes Index — October 15, 2026 snapshot
We mirror the published hermes index view for Hermes Index. Claude Opus 5.5 leads the public snapshot at 63.3%, followed by GPT-6 Astra (56.3%) and Claude Sonnet 5.5 (53.1%). We do not use these results to rank models overall.
Claude Opus 5.5
Anthropic
Hermes Agent · high where offered reasoning
Hermes Agent · $4.99 per task · provisional
GPT-6 Astra
OpenAI
Hermes Agent · high where offered reasoning
Hermes Agent · $11.61 per task
Claude Sonnet 5.5
Anthropic
Hermes Agent · high where offered reasoning
Hermes Agent · $2.82 per task · provisional
15 modelsAgenticCurrentDisplay onlyUpdated October 15, 2026 snapshot
| Rank | Model | Hermes Index | Hermes Bench | TerminalBench 4 | TerminalBench Science | SkillsBench |
|---|---|---|---|---|---|---|
| 1 | Claude Opus 5.5Anthropic | 63.31$4.99* | 76.7$0.52 | 54.0$6.27* | 52.9$12.65 | 69.6$0.51 |
| 2 | GPT-6 AstraOpenAI | 56.25$11.61 | 73.7$1.52 | 41.3$12.66 | 57.1$29.84 | 52.9$2.44 |
| 3 | Claude Sonnet 5.5Anthropic | 53.14$2.82* | 73.5$0.25 | 36.5$2.77* | 35.7$8.02 | 66.9$0.23 |
| 4 | GPT-6 SolOpenAI | 44.10$2.23 | 73.5$0.26 | 23.8$2.53 | 18.6$5.70 | 60.5$0.42 |
| 5 | Grok 4.7xAI | 39.32$10.77 | 74.4$0.69 | 12.7$23.00 | 1.4$18.41 | 68.8$0.99 · 71.9 raw |
| 6 | GLM-5.3Z.AI | 39.23$3.78 | 71.8$0.18 | 38.1$7.68 | 1.4$6.82 | 45.6$0.44 |
| 7 | Kimi K3Moonshot AI | 39.16$6.50 | 71.5$0.18 | 19.0$8.57 | 8.6$16.74 | 57.5$0.53 |
| 8 | Gemini 3.8 FlashGoogle | 38.12$3.17 | 62.2$0.64 | 23.8$5.94 | 8.6$5.15 | 57.9$0.94 · 67.7 raw |
| 9 | DeepSeek V4.1 Flash 1017DeepSeek | 36.91$0.26 | 70.7$0.019 | 17.5$0.62 | 2.9$0.38 | 56.5$0.016 · 61.1 raw |
| 10 | Qwen3.8 MaxAlibaba | 36.17$6.69 | 70.5$0.32 | 17.5$6.84 | 5.7$19.09 | 51.0$0.52 |
| 11 | GLM-5.3-FlashZ.AI | 34.95$0.50 | 68.9$0.019 | 25.4$0.60 | 2.9$1.31 | 42.6$0.058 |
| 12 | GPT-6 LunaOpenAI | 33.89$0.14 | 70.5$0.015 | 6.3$0.24 | 4.8$0.28 | 54.0$0.030 |
| 13 | Step 5 PreviewStepFun | 32.83$1.75 | 69.2$0.17 | 12.7$2.63 | 7.1$3.84 | 42.3$0.35 · 46.6 raw |
| 14 | Hy4 previewTencent | 26.13$0.48 | 58.6$0.090 | 12.7$0.70 | 1.4$0.98 | 31.8$0.16 |
| 15 | Ling 3.0 FlashInclusionAI | 21.56$0.054 | 53.8$0.004 | 0.0$0.080 | 0.0$0.12 | 32.4$0.013 |
Scores are pass@1 percentages. The line beneath each score is mean cost per task in USD. Every run uses the Hermes Agent harness with high reasoning effort where the model offers it, and TerminalBench 4 runs without its 4 GPU tasks. An asterisk marks a provisional result, a partial run or cost estimate pending a full rerun. SkillsBench shows the clean score, which drops tasks where the run found the suite's public repository, with the raw score beneath it.
How to read this leaderboard
A higher index means the model finished more tasks across the four suites inside Hermes Agent. Read the cost per task beside it: a model that scores a few points lower can cost a tenth as much. Opus 5.5 and Sonnet 5.5 carry provisional cost figures that Nous marks as pending a full rerun.
Operator receipt: 15 sourced rows are currently displayable on this page; the leading published row is Claude Opus 5.5 at 63.3%.
Honest limit: The score belongs to the model together with the Hermes Agent harness and Nous Research's grading, so it is not comparable with the same suites run in another harness or against a vendor's own terminal client. Nous also makes Hermes Agent and authors Hermes Bench. TerminalBench 4 drops its 4 GPU tasks, and some SkillsBench runs found the suite's public repository, so Nous reports a clean score that drops those tasks and keeps the raw score separately. We rely on the published table and did not rerun the models.
How we show the Hermes Index
Nous Research publishes the Hermes Index as a table of 15 models. Each score is the mean pass@1 result across four suites, Hermes Bench, TerminalBench 4, TerminalBench Science and SkillsBench, with the mean cost per task beside it. We mirror the October 15, 2026 snapshot and keep every suite column.
Every run uses the Hermes Agent harness with reasoning effort set to high where the model offers it. TerminalBench 4 runs without its 4 GPU tasks. Where a SkillsBench run found the suite's public repository, the clean score drops those tasks and is the one that counts; we show the raw score beneath it.
An asterisk marks a provisional figure: a partial run or a cost estimate that Nous says is pending a full rerun. Scores belong to the model and the harness together, so the table is display only.
The published Hermes Index snapshot places Claude Opus 5.5 first at 63.3%. The third row is 10.2 points behind. The broader top-10 range is 27.1 points, so the table still separates the published systems.
15 models are shown for Hermes Index. The benchmark falls in the Agentic category. Hermes Index is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About Hermes Index
Year
2026
Tasks
Four suites: Hermes Bench (150 tasks), TerminalBench 4 without its 4 GPU tasks, TerminalBench Science and SkillsBench
Format
Mean pass@1 score across four suites, with mean cost per task
Difficulty
Frontier agentic
Nous Research runs each model through Hermes Agent with reasoning effort set to high where the model offers it. The index is the mean of the four suite scores and the mean cost per task across the same suites. Hermes Bench is Nous's own suite; the other three are public benchmarks run in the same harness.
Freshness and provenance
Version
Hermes Index (four suites, Hermes Agent harness)
Refresh cadence
Published leaderboard snapshots
Staleness state
Current
Question availability
Aggregate results published; Hermes Bench documents its grading modes and categories
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What does the Hermes Index measure?
The Hermes Index averages each model's pass@1 score on four agentic suites, Hermes Bench, TerminalBench 4, TerminalBench Science and SkillsBench, all run inside Nous Research's Hermes Agent harness. It also reports the mean cost per task, so you can see what a score costs.
Which model leads the Hermes Index?
Claude Opus 5.5 leads the October 15, 2026 table with an index of 63.31 at $4.99 per task, a cost Nous marks as provisional. GPT-6 Astra follows at 56.25 and $11.61 per task.
Why is the Hermes Index display only?
Every score combines the model with the Hermes Agent harness, a fixed reasoning-effort setting, and Nous Research's grading. The harness maker also owns one of the four suites. We show the published table for comparison and keep it out of the weighted model-only ranking.
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 5,600+ readers.
One email each week. Unsubscribe anytime.