Benchmark Almanac

Hermes Index

Claude Opus 5.5 63.3 leads since Oct 2026 Not in model cards

Nous Research averages pass@1 and cost per task over Hermes Bench, Terminal-Bench 4.0, Terminal-Bench Science and SkillsBench, all run in Hermes Agent.

The first edition on 2026-10-06 covered 14 models.

Made by Nous Research Released Oct 2026 Composite indices All tasks public Last model added Oct 2026 Scores refreshed 2 Nov 2026
Level
Expert Specialists with years of training, often over hours of work
Scored
Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
Answers
Agent task Many steps in a terminal, browser, desktop, app, game or robot.
Leaderboard portal.nousresearch.com
18
Entries on its own leaderboard, 17 models we track

Claude Opus 5.5 leads with 63.3

Best score per entry from the benchmark's own leaderboard, read on 2 Nov 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders.

18 entries. Scroll sideways for the rest.

  1. 63.3 Claude Opus 5.5
  2. 57.1 GPT-6.1 Sol
  3. 56.3 GPT-6 Astra
  4. 53.1 Claude Sonnet 5.5
  5. 44.1 GPT-6 Sol
  6. 39.3 Grok 4.7
  7. 39.2 GLM-5.3
  8. 39.2 Kimi K3
  9. 38.1 Gemini 3.8 Flash
  10. 37.0 Claude Haiku 5.5
  11. 36.9 DeepSeek V4.1 Flash
  12. 36.2 Qwen 3.8 Max
  13. 35.0 GLM-5.3-Flash
  14. 33.9 GPT-6 Luna
  15. 32.8 Step 5 Preview
  16. 26.1 Tencent Hy4 preview
  17. 24.1 Nemotron Super 3.5
  18. 21.6 Ling-3.0-flash
  1. 1 Claude Opus 5.5 63.3
  2. 2 GPT-6.1 Sol 57.1
  3. 3 GPT-6 Astra 56.3
  4. 4 Claude Sonnet 5.5 53.1
  5. 5 GPT-6 Sol 44.1
  6. 6 Grok 4.7 39.3
  7. 7 GLM-5.3 39.2
  8. 8 Kimi K3 39.2
  9. 9 Gemini 3.8 Flash 38.1
  10. 10 Claude Haiku 5.5 37.0
  11. 11 DeepSeek V4.1 Flash 36.9
  12. 12 Qwen 3.8 Max 36.2
  13. 13 GLM-5.3-Flash 35.0
  14. 14 GPT-6 Luna 33.9
  15. 15 Step 5 Preview 32.8
  16. 16 Tencent Hy4 preview 26.1
  17. 17 Nemotron Super 3.5 24.1
  18. 18 Ling-3.0-flash 21.6
#EntryScoreSettingDateSource
1 Claude Opus 5.5 63.3 Hermes Agent Sep 2026 Official leaderboard
2 GPT-6.1 Sol 57.1 Hermes Agent Sep 2026 Official leaderboard
3 GPT-6 Astra 56.3 Hermes Agent Sep 2026 Official leaderboard
4 Claude Sonnet 5.5 53.1 Hermes Agent Sep 2026 Official leaderboard
5 GPT-6 Sol 44.1 Hermes Agent Sep 2026 Official leaderboard
6 Grok 4.7 39.3 Hermes Agent Sep 2026 Official leaderboard
7 GLM-5.3 39.2 Hermes Agent Aug 2026 Official leaderboard
8 Kimi K3 39.2 Hermes Agent Jul 2026 Official leaderboard
9 Gemini 3.8 Flash 38.1 Hermes Agent Sep 2026 Official leaderboard
10 Claude Haiku 5.5 37.0 Hermes Agent Oct 2026 Official leaderboard
11 DeepSeek V4.1 Flash 36.9 Hermes Agent Sep 2026 Official leaderboard
12 Qwen 3.8 Max 36.2 Hermes Agent Aug 2026 Official leaderboard
13 GLM-5.3-Flash 35.0 Hermes Agent Aug 2026 Official leaderboard
14 GPT-6 Luna 33.9 Hermes Agent Sep 2026 Official leaderboard
15 Step 5 Preview 32.8 Hermes Agent Sep 2026 Official leaderboard
16 Tencent Hy4 preview 26.1 Hermes Agent Aug 2026 Official leaderboard
17 Nemotron Super 3.5 24.1 Hermes Agent Official leaderboard
18 Ling-3.0-flash 21.6 Hermes Agent Aug 2026 Official leaderboard

The best score, 63.3, costs $4.99 a task

Runs on the cost frontier score higher than any run that's cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

Ant Group DeepSeek Anthropic OpenAI Cost frontier One model at each effort level
0.0 20.0 40.0 60.0 80.0 $0.01 $0.10 $1.00 $10 $100 Cost per task, log scale Claude Opus 5.5 GPT-6.1 Sol Claude Haiku 5.5 GPT-6 Luna Ling-3.0-flash 0.0 20.0 40.0 60.0 80.0 $0.01 $0.10 $1.00 $10 $100 Cost per task, log scale Claude Opus 5.5 GPT-6.1 Sol
ModelEffortHarnessScoreCost per taskOutput tokens
Claude Opus 5.5 Hermes Agent 63.3 $4.99
GPT-6.1 Sol Hermes Agent 57.1 $1.19
GPT-6 Astra Hermes Agent 56.3 $12
Claude Sonnet 5.5 Hermes Agent 53.1 $2.82
GPT-6 Sol Hermes Agent 44.1 $2.23
Grok 4.7 Hermes Agent 39.3 $11
GLM-5.3 Hermes Agent 39.2 $3.78
Kimi K3 Hermes Agent 39.2 $6.50
Gemini 3.8 Flash Hermes Agent 38.1 $3.17
Claude Haiku 5.5 Hermes Agent 37.0 $0.49
DeepSeek V4.1 Flash Hermes Agent 36.9 $0.26
Qwen 3.8 Max Hermes Agent 36.2 $6.69
GLM-5.3-Flash Hermes Agent 35.0 $0.50
GPT-6 Luna Hermes Agent 33.9 $0.14
Step 5 Preview Hermes Agent 32.8 $1.75
Tencent Hy4 preview Hermes Agent 26.1 $0.48
Ling-3.0-flash Hermes Agent 21.6 $0.05

Number one over time

Lab claims count from the day the card came out, independent runs from the model’s release.

Anthropic
Own leaderboard 1 leader
Claude Opus 5.5 63.3

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

On the benchmark’s own leaderboard
20 30 40 50 60 70 20 30 40 50 60 70

No lab model card reports it yet

We read the headline results of 479 launch posts from 51 labs, and none of them list this benchmark.

More in Other

All 41 in Other
  • Personal Superintelligence Index (Meta) In use Meta's composite of seven personal and knowledge-work evals: Agents' Last Exam, AutomationBench-AA, FocusBench, JobBench, GDP.pdf, VBT and DRACO.
  • Artificial Analysis Intelligence Index v5 In use Version 5 of Artificial Analysis's index: nine evaluations in four equal categories, most with private test sets, run three times each in its own agent harness. Gemini 4 Carbon 59.0
  • Anthropic ECI (AECI) Only in system cards Index that combines a basket of capability benchmarks into one capability score to track trends across Anthropic models. Claude Opus 5.5 169.4
  • Qwen in-house English automatic evaluation Only in system cards Models answer Qwen in-house English datasets on knowledge, comprehension, coding, math and related skills. Skills and weights change between reports. Qwen2-72B 72.9%
  • Recognition of model-written user turns (Anthropic) Only in system cards Models write synthetic user turns for real Claude Code transcripts, then judge which version is real. Score is accuracy above chance. Claude Mythos Preview 0.35
  • LiveBench Dropped Models answer monthly refreshed questions across math, coding, reasoning, language, instruction following, and data analysis. Scored automatically against objective ground truth. Claude Fable 5.1 85.7%