Benchmark Almanac

Judgemark v4

Claude Opus 4.6 90.7 leads since May 2026 Not in model cards

The model under test judges fixed creative-writing samples against low and high anchor responses. The score is how well its ratings separate writer models.

Each item has a lower anchor scored 2 and an upper anchor scored 9. Scoring uses separability metrics such as omega squared. Release date not stated on the page.

Made by EQ-Bench (Sam Paech) Released May 2026 Taste and style All tasks public Last model added Aug 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
Short answer A number, word or short string matched to a key.
Leaderboard eqbench.com
42
Entries on its own leaderboard, 40 models we track

Claude Opus 4.6 leads with 90.7 among models released by Aug 2026

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. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders.

The newest model with a score here came out in Aug 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

42 entries. Scroll sideways for the rest.

  1. 90.7 Claude Opus 4.6
  2. 87.8 GPT-5.5
  3. 84.0 Claude Opus 4.7
  4. 83.4 Gemini 3.7 Flash
  5. 82.2 Claude Sonnet 4.6
  6. 79.4 Muse Spark 1.2
  7. 78.8 Claude Opus 5
  8. 78.7 Gemini 3.1 Pro
  9. 78.0 Claude Opus 4.8
  10. 77.1 Grok 4.5
  11. 73.2 GLM-5.2
  12. 72.6 GLM-5.3
  13. 72.3 Gemma 4 31B IT
  14. 72.1 GPT-5.4
  15. 70.9 Claude Sonnet 5
  16. 68.8 Gemini 3.5 Flash
  17. 67.4 Qwen 3.8 27B
  18. 67.2 GLM-5.1
  19. 60.5 Qwen3.5-27B
  20. 58.8 Gemini 3.1 Flash-Lite
  21. 57.9 MiMo-V2.5-Pro
  22. 57.3 Kimi K2.6
  23. 54.2 Qwen3.5 397B-A17B
  24. 53.4 Qwen3.6 27B
  25. 53.0 Gemma 4 26B A4B
  26. 52.9 nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4
  27. 49.6 Grok 4.3
  28. 47.1 DeepSeek-V4-Pro
  29. 46.1 Gemini 3 Flash
  30. 45.1 Qwen 3.6 Max (Preview)
  31. 40.7 meta-models/Muse-Glimmer-30B
  32. 40.5 Qwen3.5-35B-A3B
  33. 36.8 DeepSeek-V4-Flash
  34. 32.6 Qwen 3.6 35B-A3B
  35. 32.4 Qwen3.5-9B
  36. 31.2 Qwen 3.6 Flash
  37. 30.8 DeepSeek V4 Flash 0731
  38. 28.1 Step 3.7 Flash
  39. 27.5 GPT-5.4 Nano
  40. 27.3 GPT-5.4 Mini
  41. 16.5 Mistral Small 4
  42. 15.5 gpt-oss-120b
  1. 1 Claude Opus 4.6 90.7
  2. 2 GPT-5.5 87.8
  3. 3 Claude Opus 4.7 84.0
  4. 4 Gemini 3.7 Flash 83.4
  5. 5 Claude Sonnet 4.6 82.2
  6. 6 Muse Spark 1.2 79.4
  7. 7 Claude Opus 5 78.8
  8. 8 Gemini 3.1 Pro 78.7
  9. 9 Claude Opus 4.8 78.0
  10. 10 Grok 4.5 77.1
  11. 11 GLM-5.2 73.2
  12. 12 GLM-5.3 72.6
  13. 13 Gemma 4 31B IT 72.3
  14. 14 GPT-5.4 72.1
  15. 15 Claude Sonnet 5 70.9
  16. 16 Gemini 3.5 Flash 68.8
  17. 17 Qwen 3.8 27B 67.4
  18. 18 GLM-5.1 67.2
  19. 19 Qwen3.5-27B 60.5
  20. 20 Gemini 3.1 Flash-Lite 58.8
  21. 21 MiMo-V2.5-Pro 57.9
  22. 22 Kimi K2.6 57.3
  23. 23 Qwen3.5 397B-A17B 54.2
  24. 24 Qwen3.6 27B 53.4
  25. 25 Gemma 4 26B A4B 53.0
  26. 26 nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4 52.9
  27. 27 Grok 4.3 49.6
  28. 28 DeepSeek-V4-Pro 47.1
  29. 29 Gemini 3 Flash 46.1
  30. 30 Qwen 3.6 Max (Preview) 45.1
  31. 31 meta-models/Muse-Glimmer-30B 40.7
  32. 32 Qwen3.5-35B-A3B 40.5
  33. 33 DeepSeek-V4-Flash 36.8
  34. 34 Qwen 3.6 35B-A3B 32.6
  35. 35 Qwen3.5-9B 32.4
  36. 36 Qwen 3.6 Flash 31.2
  37. 37 DeepSeek V4 Flash 0731 30.8
  38. 38 Step 3.7 Flash 28.1
  39. 39 GPT-5.4 Nano 27.5
  40. 40 GPT-5.4 Mini 27.3
  41. 41 Mistral Small 4 16.5
  42. 42 gpt-oss-120b 15.5
#EntryScoreSettingDateSource
1 Claude Opus 4.6 90.7 Feb 2026 Official leaderboard
2 GPT-5.5 87.8 Apr 2026 Official leaderboard
3 Claude Opus 4.7 84.0 Apr 2026 Official leaderboard
4 Gemini 3.7 Flash 83.4 Aug 2026 Official leaderboard
5 Claude Sonnet 4.6 82.2 Feb 2026 Official leaderboard
6 Muse Spark 1.2 79.4 Aug 2026 Official leaderboard
7 Claude Opus 5 78.8 Jul 2026 Official leaderboard
8 Gemini 3.1 Pro 78.7 Feb 2026 Official leaderboard
9 Claude Opus 4.8 78.0 May 2026 Official leaderboard
10 Grok 4.5 77.1 Jul 2026 Official leaderboard
11 GLM-5.2 73.2 Jun 2026 Official leaderboard
12 GLM-5.3 72.6 Aug 2026 Official leaderboard
13 Gemma 4 31B IT 72.3 Apr 2026 Official leaderboard
14 GPT-5.4 72.1 Mar 2026 Official leaderboard
15 Claude Sonnet 5 70.9 Jun 2026 Official leaderboard
16 Gemini 3.5 Flash 68.8 May 2026 Official leaderboard
17 Qwen 3.8 27B 67.4 Aug 2026 Official leaderboard
18 GLM-5.1 67.2 Apr 2026 Official leaderboard
19 Qwen3.5-27B 60.5 Feb 2026 Official leaderboard
20 Gemini 3.1 Flash-Lite 58.8 Mar 2026 Official leaderboard
21 MiMo-V2.5-Pro 57.9 Apr 2026 Official leaderboard
22 Kimi K2.6 57.3 Apr 2026 Official leaderboard
23 Qwen3.5 397B-A17B 54.2 Feb 2026 Official leaderboard
24 Qwen3.6 27B 53.4 Apr 2026 Official leaderboard
25 Gemma 4 26B A4B 53.0 Apr 2026 Official leaderboard
26 nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4 52.9 Official leaderboard
27 Grok 4.3 49.6 Apr 2026 Official leaderboard
28 DeepSeek-V4-Pro 47.1 Apr 2026 Official leaderboard
29 Gemini 3 Flash 46.1 Dec 2025 Official leaderboard
30 Qwen 3.6 Max (Preview) 45.1 Apr 2026 Official leaderboard
31 meta-models/Muse-Glimmer-30B 40.7 Official leaderboard
32 Qwen3.5-35B-A3B 40.5 Feb 2026 Official leaderboard
33 DeepSeek-V4-Flash 36.8 Apr 2026 Official leaderboard
34 Qwen 3.6 35B-A3B 32.6 Apr 2026 Official leaderboard
35 Qwen3.5-9B 32.4 Feb 2026 Official leaderboard
36 Qwen 3.6 Flash 31.2 Apr 2026 Official leaderboard
37 DeepSeek V4 Flash 0731 30.8 Jul 2026 Official leaderboard
38 Step 3.7 Flash 28.1 May 2026 Official leaderboard
39 GPT-5.4 Nano 27.5 Mar 2026 Official leaderboard
40 GPT-5.4 Mini 27.3 Mar 2026 Official leaderboard
41 Mistral Small 4 16.5 Mar 2026 Official leaderboard
42 gpt-oss-120b 15.5 Aug 2025 Official leaderboard

Number one over time

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

Anthropic
Own leaderboard Not tested since Aug 2026
Claude Opus 4.6 90.7

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
0.00 20.0 40.0 60.0 80.0 100.0 2026 0.00 20.0 40.0 60.0 80.0 100.0 2026

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.

Versions

Replaces Judgemark v2.1

More in Writing and taste

All 24 in Writing and taste
  • Design Bench (Moonshot) In use Models make graphic designs from direct prompts or few-shot reference images. An automated grader rates the results as Elo scores. Kimi K3.1 Pro 957
  • WritingBench Fading Models complete 1,000 real-world writing requests across 6 domains and 100 subdomains. A judge model scores each reply on five query-specific criteria. GPT-5 83.9
  • Creative Writing (MiniMax in-house) Only in system cards Models write creative pieces for real Hailuo AI user requests. Automatic assessments score the writing. MiniMax-Text-01 81.3%
  • DeepSeek Chinese functional writing evaluation Only in system cards Models answer 3,170 everyday Chinese writing prompts such as reports, emails and speeches. Pairwise judging against a baseline model gives a win rate. DeepSeek-V4-Pro 62.6%
  • Slide generation human evaluation (Z.ai) Only in system cards People compare HTML slide decks from two models on content, layout and visual quality. Metric is overall win rate. GLM-5 67.5%
  • Task-dependent output homogenization evaluation Only in system cards Models answer prompts several times. A judge scores functional diversity where it is wanted and consistency where one answer is right. Muse Spark 0.21