Benchmark Almanac

Miðeind Icelandic Grammatical Error Detection

Claude Fable 5 91.5% leads since June 2026
Not in model cards

Models decide whether an Icelandic sentence contains a grammatical error. The test uses 100 correct sentences and 100 incorrect sentences from the Icelandic Error Corpus.

Passes test 2. Miðeind reports scores for 94 named models on its archived leaderboard.

Made by
Miðeind
Released
December 2023
Kind
Reading in many languages
Level
School
Scored
Checked automatically
Answers
Multiple choice
Tasks
All public
Measures
Capability
Language
Icelandic
In and out
Text → Text
Interaction
Single turn
Items from
Another benchmark
Refresh
One fixed set
Run by
Anyone
Strongest model tested
Claude Fable 5 #12 on the capability index
Last model added
August 2026
Scores refreshed
November 20th, 2026
92
Entries on its own leaderboard, 79 models we track

Claude Fable 5 leads with 91.5% among models released by August 2026 Claude Fable 5 leads with 91.5% among models released by August 2026

Best score per entry from the benchmark's own leaderboard, read on November 20th, 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. This view shows only models we track, each at its best.

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

92 entries. Scroll sideways for the rest.

  1. 91.5 Claude Fable 5
  2. 84.5 Claude Opus 4.7
  3. 84.0 Claude Opus 4.8
  4. 82.0 Gemini 3.7 Flash
  5. 81.0 Gemini 3 Pro
  6. 80.5 Gemini 3.1 Pro
  7. 78.0 GPT-5.4
  8. 77.5 Claude Sonnet 4.6
  9. 77.5 GPT-5.5
  10. 76.5 Claude Opus 4.6
  11. 76.5 Gemini 3.6 Flash
  12. 75.5 GPT-5.6 Sol
  13. 75.0 GPT-5.1
  14. 75.0 Gemini 3.5 Flash
  15. 74.5 o1-preview
  16. 74.5 Claude Opus 4.5
  17. 74.5 GPT-5.6 Luna
  18. 73.5 Gemini 3 Flash
  19. 72.5 Claude 3.5 Sonnet
  20. 72.5 Gemini 2.5 Flash
  21. 72.5 Gemini 3.1 Flash-Lite
  22. 71.5 Gemini 2.5 Pro
  23. 71.5 GPT-5
  24. 70.0 GPT-5.6 Terra
  25. 69.5 GPT-5.2
  26. 68.0 GPT-4o
  27. 67.0 Claude 3.7 Sonnet
  28. 66.5 GPT-4.1
  29. 66.0 Grok 4.3
  30. 65.0 GPT-4o
  31. 64.0 Kimi K2.5
  32. 64.0 Gemma 4 31B IT
  33. 63.5 Claude 3 Opus
  34. 63.5 Qwen3.5 397B-A17B
  35. 63.0 Gemini 2.5 Flash-Lite
  36. 62.5 Gemini 2.0 Flash
  37. 62.5 o4-mini
  38. 62.5 DeepSeek-V4-Pro
  39. 62.0 GLM-5.2
  40. 61.5 Claude 3.5 Haiku
  41. 60.5 GPT-4 Turbo
  42. 60.5 GPT-5 mini
  43. 60.5 Claude Sonnet 4.5
  44. 60.0 Gemini 2.0 Pro
  45. 59.5 GPT-5 nano
  46. 59.0 GPT-4.1 mini
  47. 59.0 DeepSeek-R1
  48. 59.0 MiniMax-M2.5
  49. 59.0 Gemma 4 26B A4B
  50. 58.5 Gemma 3 27B
  51. 58.5 Mistral Large 3
  52. 58.0 o3-mini
  53. 57.0 GPT-4o mini
  54. 57.0 DeepSeek-V3
  55. 57.0 GPT-4.1 nano
  56. 56.5 Claude Sonnet 4
  57. 56.5 Mistral Medium 3.1
  58. 56.0 cyboghostginx/gemma-3-4b-it-Adetayo
  59. 55.5 GPT-4
  60. 55.0 DeepSeek-R1-Distill-Llama-70B
  61. 54.5 cyboghostginx/gemma-3-27b-it-Adetayo
  62. 54.0 Gemma 3 12B
  63. 53.5 NousResearch/Hermes-3-Llama-3.1-405B-FP8
  64. 53.0 GPT-4
  65. 53.0 Mistral NeMo
  66. 53.0 Llama 3.1-70B
  67. 52.5 Claude 2.1
  68. 52.5 Llama-3.1-Nemotron-70B-Instruct
  69. 52.5 Llama 3.3 70B Instruct
  70. 52.0 GPT-3.5 Turbo
  71. 52.0 Llama 3-8B
  72. 52.0 Gemma 3 4B
  73. 51.5 AI-Sweden-Models/gpt-sw3-6.7b-v2-instruct
  74. 51.5 LumiOpen/Viking-33B
  75. 51.0 AI-Sweden-Models/gpt-sw3-126m-instruct
  76. 51.0 cyboghostginx/gemma-4-31B-it-Adetayo
  77. 51.0 Qwen2-7B-Instruct
  78. 51.0 Gemma 2 9B
  79. 51.0 o1-mini
  80. 51.0 Ministral 3B
  81. 50.5 AI-Sweden-Models/gpt-sw3-1.3b-instruct
  82. 50.0 AI-Sweden-Models/gpt-sw3-20b-instruct
  83. 50.0 Llama 3.1-8B
  84. 50.0 Llama 3.2 1B
  85. 49.5 LumiOpen/Viking-13B
  86. 49.5 Gemma 3 1B
  87. 49.0 Llama 3-70B
  88. 49.0 Llama 3.2 3B
  89. 48.5 Qwen2-72B
  90. 48.0 AI-Sweden-Models/gpt-sw3-356m-instruct
  91. 48.0 Sigurdur/icebreaker
  92. 40.0 LumiOpen/Viking-7B
  1. 1 Claude Fable 5 high 91.5%
  2. 2 Claude Opus 4.7 high 84.5%
  3. 3 Claude Opus 4.8 high 84%
  4. 4 Gemini 3.7 Flash high 82%
  5. 5 Gemini 3 Pro high 81%
  6. 6 Gemini 3.1 Pro high 80.5%
  7. 7 GPT-5.4 high 78%
  8. 8 Claude Sonnet 4.6 high 77.5%
  9. 9 GPT-5.5 high 77.5%
  10. 10 Claude Opus 4.6 high 76.5%
  11. 11 Gemini 3.6 Flash high 76.5%
  12. 12 GPT-5.6 Sol high 75.5%
  13. 13 GPT-5.1 high 75%
  14. 14 Gemini 3.5 Flash high 75%
  15. 15 o1-preview on 74.5%
  16. 16 Claude Opus 4.5 high 74.5%
  17. 17 GPT-5.6 Luna high 74.5%
  18. 18 Gemini 3 Flash high 73.5%
  19. 19 Claude 3.5 Sonnet 72.5%
  20. 20 Gemini 2.5 Flash Sep 2025, high 72.5%
  21. 21 Gemini 3.1 Flash-Lite high 72.5%
  22. 22 Gemini 2.5 Pro Jun 2025, on 71.5%
  23. 23 GPT-5 high 71.5%
  24. 24 GPT-5.6 Terra high 70%
  25. 25 GPT-5.2 high 69.5%
  26. 26 GPT-4o Aug 2024 68%
  27. 27 Claude 3.7 Sonnet 67%
  28. 28 GPT-4.1 66.5%
  29. 29 Grok 4.3 on 66%
  30. 30 GPT-4o May 2024 65%
  31. 31 Kimi K2.5 on 64%
  32. 32 Gemma 4 31B IT on 64%
  33. 33 Claude 3 Opus 63.5%
  34. 34 Qwen3.5 397B-A17B on 63.5%
  35. 35 Gemini 2.5 Flash-Lite Sep 2025, high 63%
  36. 36 Gemini 2.0 Flash Feb 2025 62.5%
  37. 37 o4-mini on 62.5%
  38. 38 DeepSeek-V4-Pro on 62.5%
  39. 39 GLM-5.2 on 62%
  40. 40 Claude 3.5 Haiku 61.5%
  41. 41 GPT-4 Turbo Apr 2024 60.5%
  42. 42 GPT-5 mini high 60.5%
  43. 43 Claude Sonnet 4.5 high 60.5%
  44. 44 Gemini 2.0 Pro on 60%
  45. 45 GPT-5 nano high 59.5%
  46. 46 GPT-4.1 mini 59%
  47. 47 DeepSeek-R1 May 2025, on 59%
  48. 48 MiniMax-M2.5 on 59%
  49. 49 Gemma 4 26B A4B on 59%
  50. 50 Gemma 3 27B 58.5%
  51. 51 Mistral Large 3 58.5%
  52. 52 o3-mini on 58%
  53. 53 GPT-4o mini 57%
  54. 54 DeepSeek-V3 Mar 2025 57%
  55. 55 GPT-4.1 nano 57%
  56. 56 Claude Sonnet 4 on 56.5%
  57. 57 Mistral Medium 3.1 56.5%
  58. 58 cyboghostginx/gemma-3-4b-it-Adetayo 56%
  59. 59 GPT-4 Mar 2023 55.5%
  60. 60 DeepSeek-R1-Distill-Llama-70B on 55%
  61. 61 cyboghostginx/gemma-3-27b-it-Adetayo 54.5%
  62. 62 Gemma 3 12B 54%
  63. 63 NousResearch/Hermes-3-Llama-3.1-405B-FP8 53.5%
  64. 64 GPT-4 Jun 2023 53%
  65. 65 Mistral NeMo 53%
  66. 66 Llama 3.1-70B 53%
  67. 67 Claude 2.1 52.5%
  68. 68 Llama-3.1-Nemotron-70B-Instruct 52.5%
  69. 69 Llama 3.3 70B Instruct 52.5%
  70. 70 GPT-3.5 Turbo Jan 2024 52%
  71. 71 Llama 3-8B 52%
  72. 72 Gemma 3 4B 52%
  73. 73 AI-Sweden-Models/gpt-sw3-6.7b-v2-instruct 51.5%
  74. 74 LumiOpen/Viking-33B 51.5%
  75. 75 AI-Sweden-Models/gpt-sw3-126m-instruct 51%
  76. 76 cyboghostginx/gemma-4-31B-it-Adetayo 51%
  77. 77 Qwen2-7B-Instruct 51%
  78. 78 Gemma 2 9B 51%
  79. 79 o1-mini on 51%
  80. 80 Ministral 3B 51%
  81. 81 AI-Sweden-Models/gpt-sw3-1.3b-instruct 50.5%
  82. 82 AI-Sweden-Models/gpt-sw3-20b-instruct 50%
  83. 83 Llama 3.1-8B 50%
  84. 84 Llama 3.2 1B 50%
  85. 85 LumiOpen/Viking-13B 49.5%
  86. 86 Gemma 3 1B 49.5%
  87. 87 Llama 3-70B 49%
  88. 88 Llama 3.2 3B 49%
  89. 89 Qwen2-72B 48.5%
  90. 90 AI-Sweden-Models/gpt-sw3-356m-instruct 48%
  91. 91 Sigurdur/icebreaker 48%
  92. 92 LumiOpen/Viking-7B 40%

Source: Official Miðeind Icelandic Grammatical Error Detection leaderboard

79 models. Scroll sideways for the rest.

  1. 1 Claude Fable 5 high 91.5%
  2. 2 Claude Opus 4.7 high 84.5%
  3. 3 Claude Opus 4.8 high 84%
  4. 4 Gemini 3.7 Flash high 82%
  5. 5 Gemini 3 Pro high 81%
  6. 6 Gemini 3.1 Pro high 80.5%
  7. 7 GPT-5.4 high 78%
  8. 8 Claude Sonnet 4.6 high 77.5%
  9. 9 GPT-5.5 high 77.5%
  10. 10 Claude Opus 4.6 high 76.5%
  11. 11 Gemini 3.6 Flash high 76.5%
  12. 12 GPT-5.6 Sol high 75.5%
  13. 13 GPT-5.1 high 75%
  14. 14 Gemini 3.5 Flash high 75%
  15. 15 o1-preview on 74.5%
  16. 16 Claude Opus 4.5 high 74.5%
  17. 17 GPT-5.6 Luna high 74.5%
  18. 18 Gemini 3 Flash high 73.5%
  19. 19 Claude 3.5 Sonnet 72.5%
  20. 20 Gemini 2.5 Flash Sep 2025, high 72.5%
  21. 21 Gemini 3.1 Flash-Lite high 72.5%
  22. 22 Gemini 2.5 Pro Jun 2025, on 71.5%
  23. 23 GPT-5 high 71.5%
  24. 24 GPT-5.6 Terra high 70%
  25. 25 GPT-5.2 high 69.5%
  26. 26 GPT-4o Aug 2024 68%
  27. 27 Claude 3.7 Sonnet 67%
  28. 28 GPT-4.1 66.5%
  29. 29 Grok 4.3 on 66%
  30. 30 GPT-4o May 2024 65%
  31. 31 Kimi K2.5 on 64%
  32. 32 Gemma 4 31B IT on 64%
  33. 33 Claude 3 Opus 63.5%
  34. 34 Qwen3.5 397B-A17B on 63.5%
  35. 35 Gemini 2.5 Flash-Lite Sep 2025, high 63%
  36. 36 Gemini 2.0 Flash Feb 2025 62.5%
  37. 37 o4-mini on 62.5%
  38. 38 DeepSeek-V4-Pro on 62.5%
  39. 39 GLM-5.2 on 62%
  40. 40 Claude 3.5 Haiku 61.5%
  41. 41 GPT-4 Turbo Apr 2024 60.5%
  42. 42 GPT-5 mini high 60.5%
  43. 43 Claude Sonnet 4.5 high 60.5%
  44. 44 Gemini 2.0 Pro on 60%
  45. 45 GPT-5 nano high 59.5%
  46. 46 GPT-4.1 mini 59%
  47. 47 DeepSeek-R1 May 2025, on 59%
  48. 48 MiniMax-M2.5 on 59%
  49. 49 Gemma 4 26B A4B on 59%
  50. 50 Gemma 3 27B 58.5%
  51. 51 Mistral Large 3 58.5%
  52. 52 o3-mini on 58%
  53. 53 GPT-4o mini 57%
  54. 54 DeepSeek-V3 Mar 2025 57%
  55. 55 GPT-4.1 nano 57%
  56. 56 Claude Sonnet 4 on 56.5%
  57. 57 Mistral Medium 3.1 56.5%
  58. 58 GPT-4 Mar 2023 55.5%
  59. 59 DeepSeek-R1-Distill-Llama-70B on 55%
  60. 60 Gemma 3 12B 54%
  61. 61 GPT-4 Jun 2023 53%
  62. 62 Mistral NeMo 53%
  63. 63 Llama 3.1-70B 53%
  64. 64 Claude 2.1 52.5%
  65. 65 Llama-3.1-Nemotron-70B-Instruct 52.5%
  66. 66 Llama 3.3 70B Instruct 52.5%
  67. 67 GPT-3.5 Turbo Jan 2024 52%
  68. 68 Llama 3-8B 52%
  69. 69 Gemma 3 4B 52%
  70. 70 Qwen2-7B-Instruct 51%
  71. 71 Gemma 2 9B 51%
  72. 72 o1-mini on 51%
  73. 73 Ministral 3B 51%
  74. 74 Llama 3.1-8B 50%
  75. 75 Llama 3.2 1B 50%
  76. 76 Gemma 3 1B 49.5%
  77. 77 Llama 3-70B 49%
  78. 78 Llama 3.2 3B 49%
  79. 79 Qwen2-72B 48.5%

Source: Official Miðeind Icelandic Grammatical Error Detection leaderboard

#EntryScoreSettingDateLinkSource
1 Claude Fable 5 91.5% high Jun 2026 anthropic.com Official leaderboard
2 Claude Opus 4.7 84.5% high Apr 2026 anthropic.com Official leaderboard
3 Claude Opus 4.8 84% high May 2026 anthropic.com Official leaderboard
4 Gemini 3.7 Flash 82% high Aug 2026 ai.google.dev Official leaderboard
5 Gemini 3 Pro 81% high Nov 2025 ai.google.dev Official leaderboard
6 Gemini 3.1 Pro 80.5% high Feb 2026 ai.google.dev Official leaderboard
7 GPT-5.4 78% high Mar 2026 huggingface.co Official leaderboard
8 Claude Sonnet 4.6 77.5% high Feb 2026 anthropic.com Official leaderboard
9 GPT-5.5 77.5% high Apr 2026 developers.openai.com Official leaderboard
10 Claude Opus 4.6 76.5% high Feb 2026 anthropic.com Official leaderboard
11 Gemini 3.6 Flash 76.5% high Jul 2026 ai.google.dev Official leaderboard
12 GPT-5.6 Sol 75.5% high Jul 2026 developers.openai.com Official leaderboard
13 GPT-5.1 75% high Nov 2025 huggingface.co Official leaderboard
14 Gemini 3.5 Flash 75% high May 2026 ai.google.dev Official leaderboard
15 o1-preview 74.5% on Sep 2024 huggingface.co Official leaderboard
16 Claude Opus 4.5 74.5% high Nov 2025 huggingface.co Official leaderboard
17 GPT-5.6 Luna 74.5% high Jul 2026 developers.openai.com Official leaderboard
18 Gemini 3 Flash 73.5% high Dec 2025 ai.google.dev Official leaderboard
19 Claude 3.5 Sonnet 72.5% Jun 2024 huggingface.co Official leaderboard
20 Gemini 2.5 Flash Sep 2025 72.5% high Sep 2025 huggingface.co Official leaderboard
21 Gemini 3.1 Flash-Lite 72.5% high Mar 2026 ai.google.dev Official leaderboard
22 Gemini 2.5 Pro Jun 2025 71.5% on Jun 2025 huggingface.co Official leaderboard
23 GPT-5 71.5% high Aug 2025 huggingface.co Official leaderboard
24 GPT-5.6 Terra 70% high Jul 2026 developers.openai.com Official leaderboard
25 GPT-5.2 69.5% high Dec 2025 huggingface.co Official leaderboard
26 GPT-4o Aug 2024 68% Aug 2024 huggingface.co Official leaderboard
27 Claude 3.7 Sonnet 67% Feb 2025 huggingface.co Official leaderboard
28 GPT-4.1 66.5% Apr 2025 huggingface.co Official leaderboard
29 Grok 4.3 66% on Apr 2026 docs.x.ai Official leaderboard
30 GPT-4o May 2024 65% May 2024 huggingface.co Official leaderboard
31 Kimi K2.5 64% on Jan 2026 kimi.com Official leaderboard
32 Gemma 4 31B IT 64% on Apr 2026 huggingface.co Official leaderboard
33 Claude 3 Opus 63.5% Feb 2024 huggingface.co Official leaderboard
34 Qwen3.5 397B-A17B 63.5% on Feb 2026 huggingface.co Official leaderboard
35 Gemini 2.5 Flash-Lite Sep 2025 63% high Sep 2025 huggingface.co Official leaderboard
36 Gemini 2.0 Flash Feb 2025 62.5% Feb 2025 huggingface.co Official leaderboard
37 o4-mini 62.5% on Apr 2025 huggingface.co Official leaderboard
38 DeepSeek-V4-Pro 62.5% on Apr 2026 huggingface.co Official leaderboard
39 GLM-5.2 62% on Jun 2026 huggingface.co Official leaderboard
40 Claude 3.5 Haiku 61.5% Oct 2024 huggingface.co Official leaderboard
41 GPT-4 Turbo Apr 2024 60.5% Apr 2024 huggingface.co Official leaderboard
42 GPT-5 mini 60.5% high Aug 2025 huggingface.co Official leaderboard
43 Claude Sonnet 4.5 60.5% high Sep 2025 huggingface.co Official leaderboard
44 Gemini 2.0 Pro 60% on Feb 2025 huggingface.co Official leaderboard
45 GPT-5 nano 59.5% high Aug 2025 huggingface.co Official leaderboard
46 GPT-4.1 mini 59% Apr 2025 huggingface.co Official leaderboard
47 DeepSeek-R1 May 2025 59% on May 2025 huggingface.co Official leaderboard
48 MiniMax-M2.5 59% on Feb 2026 minimax.io Official leaderboard
49 Gemma 4 26B A4B 59% on Apr 2026 huggingface.co Official leaderboard
50 Gemma 3 27B 58.5% Mar 2025 huggingface.co Official leaderboard
51 Mistral Large 3 58.5% Dec 2025 huggingface.co Official leaderboard
52 o3-mini 58% on Jan 2025 huggingface.co Official leaderboard
53 GPT-4o mini 57% Jul 2024 huggingface.co Official leaderboard
54 DeepSeek-V3 Mar 2025 57% Mar 2025 huggingface.co Official leaderboard
55 GPT-4.1 nano 57% Apr 2025 huggingface.co Official leaderboard
56 Claude Sonnet 4 56.5% on May 2025 huggingface.co Official leaderboard
57 Mistral Medium 3.1 56.5% Aug 2025 huggingface.co Official leaderboard
58 cyboghostginx/gemma-3-4b-it-Adetayo 4B 56% huggingface.co Official leaderboard
59 GPT-4 Mar 2023 55.5% Mar 2023 huggingface.co Official leaderboard
60 DeepSeek-R1-Distill-Llama-70B 55% on Jan 2025 huggingface.co Official leaderboard
61 cyboghostginx/gemma-3-27b-it-Adetayo 27B 54.5% huggingface.co Official leaderboard
62 Gemma 3 12B 54% Mar 2025 huggingface.co Official leaderboard
63 NousResearch/Hermes-3-Llama-3.1-405B-FP8 405.9B 53.5% huggingface.co Official leaderboard
64 GPT-4 Jun 2023 53% Jun 2023 huggingface.co Official leaderboard
65 Mistral NeMo 53% Jul 2024 huggingface.co Official leaderboard
66 Llama 3.1-70B 53% Jul 2024 huggingface.co Official leaderboard
67 Claude 2.1 52.5% Nov 2023 huggingface.co Official leaderboard
68 Llama-3.1-Nemotron-70B-Instruct 52.5% Oct 2024 huggingface.co Official leaderboard
69 Llama 3.3 70B Instruct 52.5% Dec 2024 huggingface.co Official leaderboard
70 GPT-3.5 Turbo Jan 2024 52% Jan 2024 huggingface.co Official leaderboard
71 Llama 3-8B 52% Apr 2024 huggingface.co Official leaderboard
72 Gemma 3 4B 52% Mar 2025 huggingface.co Official leaderboard
73 AI-Sweden-Models/gpt-sw3-6.7b-v2-instruct 7.1B 51.5% huggingface.co Official leaderboard
74 LumiOpen/Viking-33B 33.1B 51.5% huggingface.co Official leaderboard
75 AI-Sweden-Models/gpt-sw3-126m-instruct 0.2B 51% huggingface.co Official leaderboard
76 cyboghostginx/gemma-4-31B-it-Adetayo 31B 51% huggingface.co Official leaderboard
77 Qwen2-7B-Instruct 51% Jun 2024 huggingface.co Official leaderboard
78 Gemma 2 9B 51% Jun 2024 huggingface.co Official leaderboard
79 o1-mini 51% on Sep 2024 huggingface.co Official leaderboard
80 Ministral 3B 51% Oct 2024 huggingface.co Official leaderboard
81 AI-Sweden-Models/gpt-sw3-1.3b-instruct 1.4B 50.5% huggingface.co Official leaderboard
82 AI-Sweden-Models/gpt-sw3-20b-instruct 20.9B 50% huggingface.co Official leaderboard
83 Llama 3.1-8B 50% Jul 2024 huggingface.co Official leaderboard
84 Llama 3.2 1B 50% Sep 2024 huggingface.co Official leaderboard
85 LumiOpen/Viking-13B 14B 49.5% huggingface.co Official leaderboard
86 Gemma 3 1B 49.5% Mar 2025 huggingface.co Official leaderboard
87 Llama 3-70B 49% Apr 2024 huggingface.co Official leaderboard
88 Llama 3.2 3B 49% Sep 2024 huggingface.co Official leaderboard
89 Qwen2-72B 48.5% Jun 2024 huggingface.co Official leaderboard
90 AI-Sweden-Models/gpt-sw3-356m-instruct 0.5B 48% huggingface.co Official leaderboard
91 Sigurdur/icebreaker 0.1B 48% huggingface.co Official leaderboard
92 LumiOpen/Viking-7B 7.6B 40% huggingface.co Official leaderboard
#ModelScoreSettingDateLinkSource
1 Claude Fable 5 91.5% high Jun 2026 anthropic.com Official leaderboard
2 Claude Opus 4.7 84.5% high Apr 2026 anthropic.com Official leaderboard
3 Claude Opus 4.8 84% high May 2026 anthropic.com Official leaderboard
4 Gemini 3.7 Flash 82% high Aug 2026 ai.google.dev Official leaderboard
5 Gemini 3 Pro 81% high Nov 2025 ai.google.dev Official leaderboard
6 Gemini 3.1 Pro 80.5% high Feb 2026 ai.google.dev Official leaderboard
7 GPT-5.4 78% high Mar 2026 huggingface.co Official leaderboard
8 Claude Sonnet 4.6 77.5% high Feb 2026 anthropic.com Official leaderboard
9 GPT-5.5 77.5% high Apr 2026 developers.openai.com Official leaderboard
10 Claude Opus 4.6 76.5% high Feb 2026 anthropic.com Official leaderboard
11 Gemini 3.6 Flash 76.5% high Jul 2026 ai.google.dev Official leaderboard
12 GPT-5.6 Sol 75.5% high Jul 2026 developers.openai.com Official leaderboard
13 GPT-5.1 75% high Nov 2025 huggingface.co Official leaderboard
14 Gemini 3.5 Flash 75% high May 2026 ai.google.dev Official leaderboard
15 o1-preview 74.5% on Sep 2024 huggingface.co Official leaderboard
16 Claude Opus 4.5 74.5% high Nov 2025 huggingface.co Official leaderboard
17 GPT-5.6 Luna 74.5% high Jul 2026 developers.openai.com Official leaderboard
18 Gemini 3 Flash 73.5% high Dec 2025 ai.google.dev Official leaderboard
19 Claude 3.5 Sonnet 72.5% Jun 2024 huggingface.co Official leaderboard
20 Gemini 2.5 Flash Sep 2025 72.5% high Sep 2025 huggingface.co Official leaderboard
21 Gemini 3.1 Flash-Lite 72.5% high Mar 2026 ai.google.dev Official leaderboard
22 Gemini 2.5 Pro Jun 2025 71.5% on Jun 2025 huggingface.co Official leaderboard
23 GPT-5 71.5% high Aug 2025 huggingface.co Official leaderboard
24 GPT-5.6 Terra 70% high Jul 2026 developers.openai.com Official leaderboard
25 GPT-5.2 69.5% high Dec 2025 huggingface.co Official leaderboard
26 GPT-4o Aug 2024 68% Aug 2024 huggingface.co Official leaderboard
27 Claude 3.7 Sonnet 67% Feb 2025 huggingface.co Official leaderboard
28 GPT-4.1 66.5% Apr 2025 huggingface.co Official leaderboard
29 Grok 4.3 66% on Apr 2026 docs.x.ai Official leaderboard
30 GPT-4o May 2024 65% May 2024 huggingface.co Official leaderboard
31 Kimi K2.5 64% on Jan 2026 kimi.com Official leaderboard
32 Gemma 4 31B IT 64% on Apr 2026 huggingface.co Official leaderboard
33 Claude 3 Opus 63.5% Feb 2024 huggingface.co Official leaderboard
34 Qwen3.5 397B-A17B 63.5% on Feb 2026 huggingface.co Official leaderboard
35 Gemini 2.5 Flash-Lite Sep 2025 63% high Sep 2025 huggingface.co Official leaderboard
36 Gemini 2.0 Flash Feb 2025 62.5% Feb 2025 huggingface.co Official leaderboard
37 o4-mini 62.5% on Apr 2025 huggingface.co Official leaderboard
38 DeepSeek-V4-Pro 62.5% on Apr 2026 huggingface.co Official leaderboard
39 GLM-5.2 62% on Jun 2026 huggingface.co Official leaderboard
40 Claude 3.5 Haiku 61.5% Oct 2024 huggingface.co Official leaderboard
41 GPT-4 Turbo Apr 2024 60.5% Apr 2024 huggingface.co Official leaderboard
42 GPT-5 mini 60.5% high Aug 2025 huggingface.co Official leaderboard
43 Claude Sonnet 4.5 60.5% high Sep 2025 huggingface.co Official leaderboard
44 Gemini 2.0 Pro 60% on Feb 2025 huggingface.co Official leaderboard
45 GPT-5 nano 59.5% high Aug 2025 huggingface.co Official leaderboard
46 GPT-4.1 mini 59% Apr 2025 huggingface.co Official leaderboard
47 DeepSeek-R1 May 2025 59% on May 2025 huggingface.co Official leaderboard
48 MiniMax-M2.5 59% on Feb 2026 minimax.io Official leaderboard
49 Gemma 4 26B A4B 59% on Apr 2026 huggingface.co Official leaderboard
50 Gemma 3 27B 58.5% Mar 2025 huggingface.co Official leaderboard
51 Mistral Large 3 58.5% Dec 2025 huggingface.co Official leaderboard
52 o3-mini 58% on Jan 2025 huggingface.co Official leaderboard
53 GPT-4o mini 57% Jul 2024 huggingface.co Official leaderboard
54 DeepSeek-V3 Mar 2025 57% Mar 2025 huggingface.co Official leaderboard
55 GPT-4.1 nano 57% Apr 2025 huggingface.co Official leaderboard
56 Claude Sonnet 4 56.5% on May 2025 huggingface.co Official leaderboard
57 Mistral Medium 3.1 56.5% Aug 2025 huggingface.co Official leaderboard
58 GPT-4 Mar 2023 55.5% Mar 2023 huggingface.co Official leaderboard
59 DeepSeek-R1-Distill-Llama-70B 55% on Jan 2025 huggingface.co Official leaderboard
60 Gemma 3 12B 54% Mar 2025 huggingface.co Official leaderboard
61 GPT-4 Jun 2023 53% Jun 2023 huggingface.co Official leaderboard
62 Mistral NeMo 53% Jul 2024 huggingface.co Official leaderboard
63 Llama 3.1-70B 53% Jul 2024 huggingface.co Official leaderboard
64 Claude 2.1 52.5% Nov 2023 huggingface.co Official leaderboard
65 Llama-3.1-Nemotron-70B-Instruct 52.5% Oct 2024 huggingface.co Official leaderboard
66 Llama 3.3 70B Instruct 52.5% Dec 2024 huggingface.co Official leaderboard
67 GPT-3.5 Turbo Jan 2024 52% Jan 2024 huggingface.co Official leaderboard
68 Llama 3-8B 52% Apr 2024 huggingface.co Official leaderboard
69 Gemma 3 4B 52% Mar 2025 huggingface.co Official leaderboard
70 Qwen2-7B-Instruct 51% Jun 2024 huggingface.co Official leaderboard
71 Gemma 2 9B 51% Jun 2024 huggingface.co Official leaderboard
72 o1-mini 51% on Sep 2024 huggingface.co Official leaderboard
73 Ministral 3B 51% Oct 2024 huggingface.co Official leaderboard
74 Llama 3.1-8B 50% Jul 2024 huggingface.co Official leaderboard
75 Llama 3.2 1B 50% Sep 2024 huggingface.co Official leaderboard
76 Gemma 3 1B 49.5% Mar 2025 huggingface.co Official leaderboard
77 Llama 3-70B 49% Apr 2024 huggingface.co Official leaderboard
78 Llama 3.2 3B 49% Sep 2024 huggingface.co Official leaderboard
79 Qwen2-72B 48.5% Jun 2024 huggingface.co Official leaderboard

9 models have held number one on the official leaderboard since its release in December 2023

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

OpenAI Anthropic Google DeepMind
Own leaderboard Passed 85% in Jun 2026, by Claude Fable 5

Source: Official Miðeind Icelandic Grammatical Error Detection leaderboard

Claude Fable 5 on the official leaderboard passed 85% in June 2026

That was 30 months after its release. By the same age, 51% of the benchmarks from 2023 had passed 85%. All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2023 2024 2025 2026 Near the ceiling (85%) Claude Fable 5 Claude Opus 4.7 GPT-5.1 o1-preview GPT-4o GPT-4 0% 20% 40% 60% 80% 100% 2023 2024 2025 2026 Near the ceiling (85%) Claude Fable 5 Claude Opus 4.7

Source: Official Miðeind Icelandic Grammatical Error Detection leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Multilingual

All 545 in Multilingual
  • MMMLU In use MMLU professionally translated into 14 languages. Multiple-choice accuracy. Claude Mythos Preview 92.7%
  • WMT24++ In use English-to-55-language translation of WMT24 test data. Scored with automatic metrics such as MetricX or xCOMET. K-EXAONE 90.5%
  • MMLU-ProX In use MMLU-Pro translated into up to 29 languages with expert review. Multiple-choice accuracy per language. Qwen3.7-Max 87%
  • CLIcK In use Models answer 1,995 multiple-choice questions on Korean language and culture. Questions come from official Korean exams and textbooks in eleven categories. Solar Open 2 90.7%
  • Global MMLU Lite 3.0 In use Models answer 400 human-translated MMLU questions per language in 23 languages. Version 3.0 adds Czech, Hungarian, Oriya, Slovak and Tajik and updates Italian. Accuracy. Claude Opus 5 95.7%
  • MMMB In use Models answer multiple-choice questions about images in six languages across 15 categories. The set has about 12,000 questions. LFM2.5-VL-3B 83%

Data audit

All 9 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 3 pass 2 notes
  • Pass Independent runs Scores from the official leaderboard (92 entries).
  • Pass Models scored 79 models we track, plus 13 other entries on its leaderboard.
  • Pass Newest score The newest is from August 2026, for Gemini 3.7 Flash.
  • Note Frontier models None of the top 10 has a score, which is fine for a retired test. The strongest tested is Claude Fable 5 (#12).
  • Note Headroom Saturated. The best is 91.5%, past 85% since June 2026.
Leaderboard import 3 pass
  • Pass Importer We read the board on mideind-icelandic-llm-leaderboard.static.hf.space: 81 rows for models we track and 13 other entries at the last read.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check