Benchmark Almanac

Hemingway-bench

Claude Fable 5 1,110 leads since June 2026
Not in model cards

Professional writers compare two models' answers blind on real writing prompts, from bedtime stories to business documents. The comparisons become ratings.

More than 5,000 blind pairwise judgments at launch.

Made by
Surge AI
Released
February 2026
Kind
Creative writing
Level
Expert
Scored
Graded by people
Answers
Open-ended
Tasks
Kept private
Strongest model tested
Claude Fable 5.1 #7 on the capability index
Last model added
September 2026
Scores refreshed
November 12th, 2026
47
Entries on its own leaderboard, 42 models we track

Claude Fable 5 leads with 1,110 among models released by September 2026

Best score per entry from the benchmark's own leaderboard, read on November 12th, 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 September 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

47 entries. Scroll sideways for the rest. Bars start at 700.

  1. 1,110 Claude Fable 5
  2. 1,105 Gemini 3.7 Flash
  3. 1,095 Claude Fable 5.1
  4. 1,093 Gemini 3.8 Flash
  5. 1,075 Gemini 3.6 Flash
  6. 1,073 Gemini 3.1 Pro
  7. 1,071 Gemini 3.5 Flash
  8. 1,069 Kimi K3
  9. 1,057 GPT-5.6 Sol
  10. 1,056 Gemini 3 Pro
  11. 1,053 GLM-5.3
  12. 1,052 Gemini 3 Flash
  13. 1,047 Claude Opus 5
  14. 1,046 DeepSeek-V4-Pro
  15. 1,045 Claude Opus 4.6
  16. 1,044 Claude Opus 4.8
  17. 1,042 Claude Opus 4.7
  18. 1,036 GLM-5.2
  19. 1,035 GPT-5.5
  20. 1,034 GLM-5.3-Flash
  21. 1,021 Kimi K2.6
  22. 1,016 Claude Opus 4.5
  23. 1,014 Tencent Hy4 preview
  24. 1,013 Muse Spark 1.2
  25. 1,009 Qwen 3.8 Max
  26. 1,001 Claude Sonnet 4.6
  27. 999 DeepSeek V4 Pro (preview)
  28. 994 GPT-5.2 Instant
  29. 990 Qwen3.5-Plus
  30. 990 Gemini 3.5 Flash-Lite
  31. 988 Kimi K2.5
  32. 985 GPT-5.4
  33. 985 Muse Spark 1.1
  34. 983 Grok 4.6
  35. 981 DeepSeek-V4-Flash-Vision-Exp
  36. 972 Grok 4.5
  37. 971 Tencent Hy3
  38. 969 DeepSeek V4 Flash (preview)
  39. 954 GPT-5.2
  40. 948 Qwen3-Max
  41. 929 Qwen3.8 Flash
  42. 916 Grok 4.1 Fast
  43. 913 Muse Glimmer
  44. 888 Nemotron Lightning 3.5 30B A3B
  45. 883 Kimi K2 Instruct
  46. 803 Llama 4 Maverick
  47. 749 Nova 2 Pro (No reasoning)

Bars start at 700.

  1. 1 Claude Fable 5 adaptive/high 1,110
  2. 2 Gemini 3.7 Flash medium 1,105
  3. 3 Claude Fable 5.1 adaptive/high 1,095
  4. 4 Gemini 3.8 Flash medium 1,093
  5. 5 Gemini 3.6 Flash medium 1,075
  6. 6 Gemini 3.1 Pro high 1,073
  7. 7 Gemini 3.5 Flash medium 1,071
  8. 8 Kimi K3 max 1,069
  9. 9 GPT-5.6 Sol medium 1,057
  10. 10 Gemini 3 Pro high 1,056
  11. 11 GLM-5.3 max 1,053
  12. 12 Gemini 3 Flash high 1,052
  13. 13 Claude Opus 5 adaptive/high 1,047
  14. 14 DeepSeek-V4-Pro high 1,046
  15. 15 Claude Opus 4.6 high 1,045
  16. 16 Claude Opus 4.8 adaptive/default/high 1,044
  17. 17 Claude Opus 4.7 high 1,042
  18. 18 GLM-5.2 auto 1,036
  19. 19 GPT-5.5 medium 1,035
  20. 20 GLM-5.3-Flash max 1,034
  21. 21 Kimi K2.6 thinking on 1,021
  22. 22 Claude Opus 4.5 high 1,016
  23. 23 Tencent Hy4 preview high 1,014
  24. 24 Muse Spark 1.2 auto 1,013
  25. 25 Qwen 3.8 Max xhigh 1,009
  26. 26 Claude Sonnet 4.6 high 1,001
  27. 27 DeepSeek V4 Pro (preview) high 999
  28. 28 GPT-5.2 Instant 994
  29. 29 Qwen3.5-Plus thinking on 990
  30. 30 Gemini 3.5 Flash-Lite minimal 990
  31. 31 Kimi K2.5 thinking on 988
  32. 32 GPT-5.4 no reasoning 985
  33. 33 Muse Spark 1.1 medium 985
  34. 34 Grok 4.6 high 983
  35. 35 DeepSeek-V4-Flash-Vision-Exp 981
  36. 36 Grok 4.5 high 972
  37. 37 Tencent Hy3 no reasoning 971
  38. 38 DeepSeek V4 Flash (preview) high 969
  39. 39 GPT-5.2 no reasoning 954
  40. 40 Qwen3-Max no reasoning 948
  41. 41 Qwen3.8 Flash xhigh 929
  42. 42 Grok 4.1 Fast thinking on 916
  43. 43 Muse Glimmer high 913
  44. 44 Nemotron Lightning 3.5 30B A3B thinking on 888
  45. 45 Kimi K2 Instruct 883
  46. 46 Llama 4 Maverick 803
  47. 47 Nova 2 Pro (No reasoning) 749

Source: Official Hemingway-bench leaderboard

#EntryScoreSettingDateSource
1 Claude Fable 5 1,110 adaptive/high Jun 2026 Official leaderboard
2 Gemini 3.7 Flash 1,105 medium Aug 2026 Official leaderboard
3 Claude Fable 5.1 1,095 adaptive/high Sep 2026 Official leaderboard
4 Gemini 3.8 Flash 1,093 medium Sep 2026 Official leaderboard
5 Gemini 3.6 Flash 1,075 medium Jul 2026 Official leaderboard
6 Gemini 3.1 Pro 1,073 high Feb 2026 Official leaderboard
7 Gemini 3.5 Flash 1,071 medium May 2026 Official leaderboard
8 Kimi K3 1,069 max Jul 2026 Official leaderboard
9 GPT-5.6 Sol 1,057 medium Jul 2026 Official leaderboard
10 Gemini 3 Pro 1,056 high Nov 2025 Official leaderboard
11 GLM-5.3 1,053 max Aug 2026 Official leaderboard
12 Gemini 3 Flash 1,052 high Dec 2025 Official leaderboard
13 Claude Opus 5 1,047 adaptive/high Jul 2026 Official leaderboard
14 DeepSeek-V4-Pro 1,046 high Apr 2026 Official leaderboard
15 Claude Opus 4.6 1,045 high Feb 2026 Official leaderboard
16 Claude Opus 4.8 1,044 adaptive/default/high May 2026 Official leaderboard
17 Claude Opus 4.7 1,042 high Apr 2026 Official leaderboard
18 GLM-5.2 1,036 auto Jun 2026 Official leaderboard
19 GPT-5.5 1,035 medium Apr 2026 Official leaderboard
20 GLM-5.3-Flash 1,034 max Aug 2026 Official leaderboard
21 Kimi K2.6 1,021 thinking on Apr 2026 Official leaderboard
22 Claude Opus 4.5 1,016 high Nov 2025 Official leaderboard
23 Tencent Hy4 preview 1,014 high Aug 2026 Official leaderboard
24 Muse Spark 1.2 1,013 auto Aug 2026 Official leaderboard
25 Qwen 3.8 Max 1,009 xhigh Aug 2026 Official leaderboard
26 Claude Sonnet 4.6 1,001 high Feb 2026 Official leaderboard
27 DeepSeek V4 Pro (preview) 999 high Official leaderboard
28 GPT-5.2 Instant 994 Dec 2025 Official leaderboard
29 Qwen3.5-Plus 990 thinking on Feb 2026 Official leaderboard
30 Gemini 3.5 Flash-Lite 990 minimal Jul 2026 Official leaderboard
31 Kimi K2.5 988 thinking on Jan 2026 Official leaderboard
32 GPT-5.4 985 no reasoning Mar 2026 Official leaderboard
33 Muse Spark 1.1 985 medium Jul 2026 Official leaderboard
34 Grok 4.6 983 high Aug 2026 Official leaderboard
35 DeepSeek-V4-Flash-Vision-Exp 981 Aug 2026 Official leaderboard
36 Grok 4.5 972 high Jul 2026 Official leaderboard
37 Tencent Hy3 971 no reasoning Jul 2026 Official leaderboard
38 DeepSeek V4 Flash (preview) 969 high Official leaderboard
39 GPT-5.2 954 no reasoning Dec 2025 Official leaderboard
40 Qwen3-Max 948 no reasoning Sep 2025 Official leaderboard
41 Qwen3.8 Flash 929 xhigh Aug 2026 Official leaderboard
42 Grok 4.1 Fast 916 thinking on Nov 2025 Official leaderboard
43 Muse Glimmer 913 high Aug 2026 Official leaderboard
44 Nemotron Lightning 3.5 30B A3B 888 thinking on Official leaderboard
45 Kimi K2 Instruct 883 Official leaderboard
46 Llama 4 Maverick 803 Apr 2025 Official leaderboard
47 Nova 2 Pro (No reasoning) 749 Official leaderboard

3 models have held number one on the official leaderboard since its release in February 2026

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

Google DeepMind Anthropic
Own leaderboard Not tested since Sep 2026
Claude Fable 5 1,110

Source: Official Hemingway-bench leaderboard

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
800 900 1,000 1,100 1,200 2025 2026 800 900 1,000 1,100 1,200 2025 2026

Source: Official Hemingway-bench leaderboard

No lab model card reports it yet

We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark.

More in Writing and taste

All 30 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