Benchmark Almanac

Miðeind Icelandic Case Declension

GPT-5.5 98% leads since April 2026
Not in model cards

Models decline Icelandic adjective and noun pairs in four cases and two numbers. The public test has 2,400 forms for 300 phrases.

Passes test 2. Miðeind reports scores for 93 named models on its archived leaderboard. The archive scores individual forms with one example. The 2023 announcement counted entire declensions.

Made by
Miðeind
Released
November 2023
Kind
Reasoning in many languages
Level
School
Scored
Checked automatically
Answers
Short answer
Tasks
All public
Measures
Capability
Language
Icelandic
In and out
Text → Text
Interaction
Single turn
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
91
Entries on its own leaderboard, 79 models we track

GPT-5.5 leads with 98% among models released by August 2026 GPT-5.5 leads with 98% 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:

91 entries. Scroll sideways for the rest.

  1. 98.0 GPT-5.5
  2. 97.8 Claude Fable 5
  3. 97.4 Claude Opus 4.8
  4. 97.3 Gemini 3.7 Flash
  5. 97.3 Gemini 3 Pro
  6. 97.1 Claude Opus 4.7
  7. 97.0 GPT-5.6 Sol
  8. 96.9 Gemini 3.5 Flash
  9. 96.7 Gemini 3.1 Pro
  10. 96.7 GPT-5.4
  11. 96.5 cyboghostginx/gemma-3-27b-it-Adetayo
  12. 96.5 Gemini 3.6 Flash
  13. 95.7 Gemini 3 Flash
  14. 95.2 GPT-5
  15. 94.8 GPT-5.1
  16. 94.7 cyboghostginx/gemma-4-31B-it-Adetayo
  17. 94.5 GPT-5.6 Luna
  18. 93.9 cyboghostginx/gemma-3-4b-it-Adetayo
  19. 92.7 Gemini 3.1 Flash-Lite
  20. 92.5 GPT-5.6 Terra
  21. 92.2 DeepSeek-V4-Pro
  22. 92.1 Claude Opus 4.6
  23. 91.3 GPT-5.2
  24. 90.8 Gemini 2.5 Pro
  25. 89.7 Gemini 2.0 Pro
  26. 89.7 Claude Opus 4.5
  27. 89.2 Claude 3.5 Sonnet
  28. 88.3 Claude 3.7 Sonnet
  29. 87.8 GPT-4o
  30. 87.8 GPT-4o
  31. 86.6 Gemini 2.0 Flash
  32. 84.6 GPT-4.1
  33. 84.4 o1-preview
  34. 84.2 Claude 3 Opus
  35. 82.5 Gemini 2.5 Flash
  36. 82.4 Grok 4.3
  37. 81.2 Kimi K2.5
  38. 80.4 Claude Sonnet 4.6
  39. 80.0 GLM-5.2
  40. 79.5 Claude Sonnet 4.5
  41. 77.7 GPT-5 mini
  42. 77.3 DeepSeek-V3
  43. 76.6 GPT-4 Turbo
  44. 74.1 Gemini 2.5 Flash-Lite
  45. 74.1 Claude Sonnet 4
  46. 73.4 Mistral Large 3
  47. 73.1 o4-mini
  48. 70.6 Claude 3.5 Haiku
  49. 69.6 GPT-4.1 mini
  50. 68.5 DeepSeek-R1
  51. 64.8 Gemma 4 31B IT
  52. 63.9 GPT-4o mini
  53. 61.8 NousResearch/Hermes-3-Llama-3.1-405B-FP8
  54. 59.6 MiniMax-M2.5
  55. 55.2 Claude 2.1
  56. 54.6 GPT-4
  57. 53.6 Gemma 4 26B A4B
  58. 53.1 Mistral Medium 3.1
  59. 50.7 GPT-4
  60. 50.4 Gemma 3 27B
  61. 48.2 Llama 3.3 70B Instruct
  62. 47.7 GPT-5 nano
  63. 46.3 Llama 3.1-70B
  64. 46.0 o3-mini
  65. 46.0 Llama-3.1-Nemotron-70B-Instruct
  66. 39.9 GPT-4.1 nano
  67. 39.1 GPT-3.5 Turbo
  68. 38.2 Qwen3.5 397B-A17B
  69. 36.5 o1-mini
  70. 32.9 Gemma 3 12B
  71. 27.2 Mistral NeMo
  72. 23.8 Llama 3-70B
  73. 21.5 LumiOpen/Viking-13B
  74. 21.5 DeepSeek-R1-Distill-Llama-70B
  75. 18.8 Llama 3.1-8B
  76. 18.0 Ministral 3B
  77. 17.7 AI-Sweden-Models/gpt-sw3-20b-instruct
  78. 17.1 Gemma 3 4B
  79. 16.9 LumiOpen/Viking-33B
  80. 15.8 LumiOpen/Viking-7B
  81. 14.1 AI-Sweden-Models/gpt-sw3-356m-instruct
  82. 13.5 AI-Sweden-Models/gpt-sw3-6.7b-v2-instruct
  83. 12.8 AI-Sweden-Models/gpt-sw3-126m-instruct
  84. 11.0 Gemma 2 9B
  85. 10.3 AI-Sweden-Models/gpt-sw3-1.3b-instruct
  86. 10.2 Llama 3.2 3B
  87. 9.71 Llama 3.2 1B
  88. 7.17 Llama 3-8B
  89. 4.46 Qwen2-72B
  90. 3.25 Qwen2-7B-Instruct
  91. 0.83 Gemma 3 1B
  1. 1 GPT-5.5 high 98%
  2. 2 Claude Fable 5 high 97.8%
  3. 3 Claude Opus 4.8 high 97.4%
  4. 4 Gemini 3.7 Flash high 97.3%
  5. 5 Gemini 3 Pro high 97.3%
  6. 6 Claude Opus 4.7 high 97.1%
  7. 7 GPT-5.6 Sol high 97%
  8. 8 Gemini 3.5 Flash high 96.9%
  9. 9 Gemini 3.1 Pro high 96.7%
  10. 10 GPT-5.4 high 96.7%
  11. 11 cyboghostginx/gemma-3-27b-it-Adetayo 96.5%
  12. 12 Gemini 3.6 Flash high 96.5%
  13. 13 Gemini 3 Flash high 95.7%
  14. 14 GPT-5 high 95.2%
  15. 15 GPT-5.1 high 94.8%
  16. 16 cyboghostginx/gemma-4-31B-it-Adetayo 94.7%
  17. 17 GPT-5.6 Luna high 94.5%
  18. 18 cyboghostginx/gemma-3-4b-it-Adetayo 93.9%
  19. 19 Gemini 3.1 Flash-Lite high 92.7%
  20. 20 GPT-5.6 Terra high 92.5%
  21. 21 DeepSeek-V4-Pro on 92.2%
  22. 22 Claude Opus 4.6 high 92.1%
  23. 23 GPT-5.2 high 91.3%
  24. 24 Gemini 2.5 Pro Jun 2025, on 90.8%
  25. 25 Gemini 2.0 Pro on 89.7%
  26. 26 Claude Opus 4.5 high 89.7%
  27. 27 Claude 3.5 Sonnet 89.2%
  28. 28 Claude 3.7 Sonnet 88.3%
  29. 29 GPT-4o May 2024 87.8%
  30. 30 GPT-4o Aug 2024 87.8%
  31. 31 Gemini 2.0 Flash Feb 2025 86.6%
  32. 32 GPT-4.1 84.6%
  33. 33 o1-preview on 84.4%
  34. 34 Claude 3 Opus 84.2%
  35. 35 Gemini 2.5 Flash Sep 2025, high 82.5%
  36. 36 Grok 4.3 on 82.4%
  37. 37 Kimi K2.5 on 81.2%
  38. 38 Claude Sonnet 4.6 high 80.4%
  39. 39 GLM-5.2 on 80%
  40. 40 Claude Sonnet 4.5 high 79.5%
  41. 41 GPT-5 mini high 77.7%
  42. 42 DeepSeek-V3 Mar 2025 77.3%
  43. 43 GPT-4 Turbo Apr 2024 76.6%
  44. 44 Gemini 2.5 Flash-Lite Sep 2025, high 74.1%
  45. 45 Claude Sonnet 4 on 74.1%
  46. 46 Mistral Large 3 73.4%
  47. 47 o4-mini on 73.1%
  48. 48 Claude 3.5 Haiku 70.6%
  49. 49 GPT-4.1 mini 69.6%
  50. 50 DeepSeek-R1 May 2025, on 68.5%
  51. 51 Gemma 4 31B IT on 64.8%
  52. 52 GPT-4o mini 63.9%
  53. 53 NousResearch/Hermes-3-Llama-3.1-405B-FP8 61.8%
  54. 54 MiniMax-M2.5 on 59.6%
  55. 55 Claude 2.1 55.2%
  56. 56 GPT-4 Mar 2023 54.6%
  57. 57 Gemma 4 26B A4B on 53.6%
  58. 58 Mistral Medium 3.1 53.1%
  59. 59 GPT-4 Jun 2023 50.7%
  60. 60 Gemma 3 27B 50.4%
  61. 61 Llama 3.3 70B Instruct 48.2%
  62. 62 GPT-5 nano high 47.7%
  63. 63 Llama 3.1-70B 46.3%
  64. 64 o3-mini on 46%
  65. 65 Llama-3.1-Nemotron-70B-Instruct 46%
  66. 66 GPT-4.1 nano 39.9%
  67. 67 GPT-3.5 Turbo Jan 2024 39.1%
  68. 68 Qwen3.5 397B-A17B on 38.2%
  69. 69 o1-mini on 36.5%
  70. 70 Gemma 3 12B 32.9%
  71. 71 Mistral NeMo 27.2%
  72. 72 Llama 3-70B 23.8%
  73. 73 LumiOpen/Viking-13B 21.5%
  74. 74 DeepSeek-R1-Distill-Llama-70B on 21.5%
  75. 75 Llama 3.1-8B 18.8%
  76. 76 Ministral 3B 18%
  77. 77 AI-Sweden-Models/gpt-sw3-20b-instruct 17.7%
  78. 78 Gemma 3 4B 17.1%
  79. 79 LumiOpen/Viking-33B 16.9%
  80. 80 LumiOpen/Viking-7B 15.8%
  81. 81 AI-Sweden-Models/gpt-sw3-356m-instruct 14.1%
  82. 82 AI-Sweden-Models/gpt-sw3-6.7b-v2-instruct 13.5%
  83. 83 AI-Sweden-Models/gpt-sw3-126m-instruct 12.8%
  84. 84 Gemma 2 9B 11%
  85. 85 AI-Sweden-Models/gpt-sw3-1.3b-instruct 10.3%
  86. 86 Llama 3.2 3B 10.2%
  87. 87 Llama 3.2 1B 9.71%
  88. 88 Llama 3-8B 7.17%
  89. 89 Qwen2-72B 4.46%
  90. 90 Qwen2-7B-Instruct 3.25%
  91. 91 Gemma 3 1B 0.83%

Source: Official Miðeind Icelandic Case Declension leaderboard

79 models. Scroll sideways for the rest.

  1. 1 GPT-5.5 high 98%
  2. 2 Claude Fable 5 high 97.8%
  3. 3 Claude Opus 4.8 high 97.4%
  4. 4 Gemini 3.7 Flash high 97.3%
  5. 5 Gemini 3 Pro high 97.3%
  6. 6 Claude Opus 4.7 high 97.1%
  7. 7 GPT-5.6 Sol high 97%
  8. 8 Gemini 3.5 Flash high 96.9%
  9. 9 Gemini 3.1 Pro high 96.7%
  10. 10 GPT-5.4 high 96.7%
  11. 11 Gemini 3.6 Flash high 96.5%
  12. 12 Gemini 3 Flash high 95.7%
  13. 13 GPT-5 high 95.2%
  14. 14 GPT-5.1 high 94.8%
  15. 15 GPT-5.6 Luna high 94.5%
  16. 16 Gemini 3.1 Flash-Lite high 92.7%
  17. 17 GPT-5.6 Terra high 92.5%
  18. 18 DeepSeek-V4-Pro on 92.2%
  19. 19 Claude Opus 4.6 high 92.1%
  20. 20 GPT-5.2 high 91.3%
  21. 21 Gemini 2.5 Pro Jun 2025, on 90.8%
  22. 22 Gemini 2.0 Pro on 89.7%
  23. 23 Claude Opus 4.5 high 89.7%
  24. 24 Claude 3.5 Sonnet 89.2%
  25. 25 Claude 3.7 Sonnet 88.3%
  26. 26 GPT-4o May 2024 87.8%
  27. 27 GPT-4o Aug 2024 87.8%
  28. 28 Gemini 2.0 Flash Feb 2025 86.6%
  29. 29 GPT-4.1 84.6%
  30. 30 o1-preview on 84.4%
  31. 31 Claude 3 Opus 84.2%
  32. 32 Gemini 2.5 Flash Sep 2025, high 82.5%
  33. 33 Grok 4.3 on 82.4%
  34. 34 Kimi K2.5 on 81.2%
  35. 35 Claude Sonnet 4.6 high 80.4%
  36. 36 GLM-5.2 on 80%
  37. 37 Claude Sonnet 4.5 high 79.5%
  38. 38 GPT-5 mini high 77.7%
  39. 39 DeepSeek-V3 Mar 2025 77.3%
  40. 40 GPT-4 Turbo Apr 2024 76.6%
  41. 41 Gemini 2.5 Flash-Lite Sep 2025, high 74.1%
  42. 42 Claude Sonnet 4 on 74.1%
  43. 43 Mistral Large 3 73.4%
  44. 44 o4-mini on 73.1%
  45. 45 Claude 3.5 Haiku 70.6%
  46. 46 GPT-4.1 mini 69.6%
  47. 47 DeepSeek-R1 May 2025, on 68.5%
  48. 48 Gemma 4 31B IT on 64.8%
  49. 49 GPT-4o mini 63.9%
  50. 50 MiniMax-M2.5 on 59.6%
  51. 51 Claude 2.1 55.2%
  52. 52 GPT-4 Mar 2023 54.6%
  53. 53 Gemma 4 26B A4B on 53.6%
  54. 54 Mistral Medium 3.1 53.1%
  55. 55 GPT-4 Jun 2023 50.7%
  56. 56 Gemma 3 27B 50.4%
  57. 57 Llama 3.3 70B Instruct 48.2%
  58. 58 GPT-5 nano high 47.7%
  59. 59 Llama 3.1-70B 46.3%
  60. 60 o3-mini on 46%
  61. 61 Llama-3.1-Nemotron-70B-Instruct 46%
  62. 62 GPT-4.1 nano 39.9%
  63. 63 GPT-3.5 Turbo Jan 2024 39.1%
  64. 64 Qwen3.5 397B-A17B on 38.2%
  65. 65 o1-mini on 36.5%
  66. 66 Gemma 3 12B 32.9%
  67. 67 Mistral NeMo 27.2%
  68. 68 Llama 3-70B 23.8%
  69. 69 DeepSeek-R1-Distill-Llama-70B on 21.5%
  70. 70 Llama 3.1-8B 18.8%
  71. 71 Ministral 3B 18%
  72. 72 Gemma 3 4B 17.1%
  73. 73 Gemma 2 9B 11%
  74. 74 Llama 3.2 3B 10.2%
  75. 75 Llama 3.2 1B 9.71%
  76. 76 Llama 3-8B 7.17%
  77. 77 Qwen2-72B 4.46%
  78. 78 Qwen2-7B-Instruct 3.25%
  79. 79 Gemma 3 1B 0.83%

Source: Official Miðeind Icelandic Case Declension leaderboard

#EntryScoreSettingDateLinkSource
1 GPT-5.5 98% high Apr 2026 developers.openai.com Official leaderboard
2 Claude Fable 5 97.8% high Jun 2026 anthropic.com Official leaderboard
3 Claude Opus 4.8 97.4% high May 2026 anthropic.com Official leaderboard
4 Gemini 3.7 Flash 97.3% high Aug 2026 ai.google.dev Official leaderboard
5 Gemini 3 Pro 97.3% high Nov 2025 ai.google.dev Official leaderboard
6 Claude Opus 4.7 97.1% high Apr 2026 anthropic.com Official leaderboard
7 GPT-5.6 Sol 97% high Jul 2026 developers.openai.com Official leaderboard
8 Gemini 3.5 Flash 96.9% high May 2026 ai.google.dev Official leaderboard
9 Gemini 3.1 Pro 96.7% high Feb 2026 ai.google.dev Official leaderboard
10 GPT-5.4 96.7% high Mar 2026 huggingface.co Official leaderboard
11 cyboghostginx/gemma-3-27b-it-Adetayo 27B 96.5% huggingface.co Official leaderboard
12 Gemini 3.6 Flash 96.5% high Jul 2026 ai.google.dev Official leaderboard
13 Gemini 3 Flash 95.7% high Dec 2025 ai.google.dev Official leaderboard
14 GPT-5 95.2% high Aug 2025 huggingface.co Official leaderboard
15 GPT-5.1 94.8% high Nov 2025 huggingface.co Official leaderboard
16 cyboghostginx/gemma-4-31B-it-Adetayo 31B 94.7% huggingface.co Official leaderboard
17 GPT-5.6 Luna 94.5% high Jul 2026 developers.openai.com Official leaderboard
18 cyboghostginx/gemma-3-4b-it-Adetayo 4B 93.9% huggingface.co Official leaderboard
19 Gemini 3.1 Flash-Lite 92.7% high Mar 2026 ai.google.dev Official leaderboard
20 GPT-5.6 Terra 92.5% high Jul 2026 developers.openai.com Official leaderboard
21 DeepSeek-V4-Pro 92.2% on Apr 2026 huggingface.co Official leaderboard
22 Claude Opus 4.6 92.1% high Feb 2026 anthropic.com Official leaderboard
23 GPT-5.2 91.3% high Dec 2025 huggingface.co Official leaderboard
24 Gemini 2.5 Pro Jun 2025 90.8% on Jun 2025 huggingface.co Official leaderboard
25 Gemini 2.0 Pro 89.7% on Feb 2025 huggingface.co Official leaderboard
26 Claude Opus 4.5 89.7% high Nov 2025 huggingface.co Official leaderboard
27 Claude 3.5 Sonnet 89.2% Jun 2024 huggingface.co Official leaderboard
28 Claude 3.7 Sonnet 88.3% Feb 2025 huggingface.co Official leaderboard
29 GPT-4o May 2024 87.8% May 2024 huggingface.co Official leaderboard
30 GPT-4o Aug 2024 87.8% Aug 2024 huggingface.co Official leaderboard
31 Gemini 2.0 Flash Feb 2025 86.6% Feb 2025 huggingface.co Official leaderboard
32 GPT-4.1 84.6% Apr 2025 huggingface.co Official leaderboard
33 o1-preview 84.4% on Sep 2024 huggingface.co Official leaderboard
34 Claude 3 Opus 84.2% Feb 2024 huggingface.co Official leaderboard
35 Gemini 2.5 Flash Sep 2025 82.5% high Sep 2025 huggingface.co Official leaderboard
36 Grok 4.3 82.4% on Apr 2026 docs.x.ai Official leaderboard
37 Kimi K2.5 81.2% on Jan 2026 kimi.com Official leaderboard
38 Claude Sonnet 4.6 80.4% high Feb 2026 anthropic.com Official leaderboard
39 GLM-5.2 80% on Jun 2026 huggingface.co Official leaderboard
40 Claude Sonnet 4.5 79.5% high Sep 2025 huggingface.co Official leaderboard
41 GPT-5 mini 77.7% high Aug 2025 huggingface.co Official leaderboard
42 DeepSeek-V3 Mar 2025 77.3% Mar 2025 huggingface.co Official leaderboard
43 GPT-4 Turbo Apr 2024 76.6% Apr 2024 huggingface.co Official leaderboard
44 Gemini 2.5 Flash-Lite Sep 2025 74.1% high Sep 2025 huggingface.co Official leaderboard
45 Claude Sonnet 4 74.1% on May 2025 huggingface.co Official leaderboard
46 Mistral Large 3 73.4% Dec 2025 huggingface.co Official leaderboard
47 o4-mini 73.1% on Apr 2025 huggingface.co Official leaderboard
48 Claude 3.5 Haiku 70.6% Oct 2024 huggingface.co Official leaderboard
49 GPT-4.1 mini 69.6% Apr 2025 huggingface.co Official leaderboard
50 DeepSeek-R1 May 2025 68.5% on May 2025 huggingface.co Official leaderboard
51 Gemma 4 31B IT 64.8% on Apr 2026 huggingface.co Official leaderboard
52 GPT-4o mini 63.9% Jul 2024 huggingface.co Official leaderboard
53 NousResearch/Hermes-3-Llama-3.1-405B-FP8 405.9B 61.8% huggingface.co Official leaderboard
54 MiniMax-M2.5 59.6% on Feb 2026 minimax.io Official leaderboard
55 Claude 2.1 55.2% Nov 2023 huggingface.co Official leaderboard
56 GPT-4 Mar 2023 54.6% Mar 2023 huggingface.co Official leaderboard
57 Gemma 4 26B A4B 53.6% on Apr 2026 huggingface.co Official leaderboard
58 Mistral Medium 3.1 53.1% Aug 2025 huggingface.co Official leaderboard
59 GPT-4 Jun 2023 50.7% Jun 2023 huggingface.co Official leaderboard
60 Gemma 3 27B 50.4% Mar 2025 huggingface.co Official leaderboard
61 Llama 3.3 70B Instruct 48.2% Dec 2024 huggingface.co Official leaderboard
62 GPT-5 nano 47.7% high Aug 2025 huggingface.co Official leaderboard
63 Llama 3.1-70B 46.3% Jul 2024 huggingface.co Official leaderboard
64 o3-mini 46% on Jan 2025 huggingface.co Official leaderboard
65 Llama-3.1-Nemotron-70B-Instruct 46% Oct 2024 huggingface.co Official leaderboard
66 GPT-4.1 nano 39.9% Apr 2025 huggingface.co Official leaderboard
67 GPT-3.5 Turbo Jan 2024 39.1% Jan 2024 huggingface.co Official leaderboard
68 Qwen3.5 397B-A17B 38.2% on Feb 2026 huggingface.co Official leaderboard
69 o1-mini 36.5% on Sep 2024 huggingface.co Official leaderboard
70 Gemma 3 12B 32.9% Mar 2025 huggingface.co Official leaderboard
71 Mistral NeMo 27.2% Jul 2024 huggingface.co Official leaderboard
72 Llama 3-70B 23.8% Apr 2024 huggingface.co Official leaderboard
73 LumiOpen/Viking-13B 14B 21.5% huggingface.co Official leaderboard
74 DeepSeek-R1-Distill-Llama-70B 21.5% on Jan 2025 huggingface.co Official leaderboard
75 Llama 3.1-8B 18.8% Jul 2024 huggingface.co Official leaderboard
76 Ministral 3B 18% Oct 2024 huggingface.co Official leaderboard
77 AI-Sweden-Models/gpt-sw3-20b-instruct 20.9B 17.7% huggingface.co Official leaderboard
78 Gemma 3 4B 17.1% Mar 2025 huggingface.co Official leaderboard
79 LumiOpen/Viking-33B 33.1B 16.9% huggingface.co Official leaderboard
80 LumiOpen/Viking-7B 7.6B 15.8% huggingface.co Official leaderboard
81 AI-Sweden-Models/gpt-sw3-356m-instruct 0.5B 14.1% huggingface.co Official leaderboard
82 AI-Sweden-Models/gpt-sw3-6.7b-v2-instruct 7.1B 13.5% huggingface.co Official leaderboard
83 AI-Sweden-Models/gpt-sw3-126m-instruct 0.2B 12.8% huggingface.co Official leaderboard
84 Gemma 2 9B 11% Jun 2024 huggingface.co Official leaderboard
85 AI-Sweden-Models/gpt-sw3-1.3b-instruct 1.4B 10.3% huggingface.co Official leaderboard
86 Llama 3.2 3B 10.2% Sep 2024 huggingface.co Official leaderboard
87 Llama 3.2 1B 9.71% Sep 2024 huggingface.co Official leaderboard
88 Llama 3-8B 7.17% Apr 2024 huggingface.co Official leaderboard
89 Qwen2-72B 4.46% Jun 2024 huggingface.co Official leaderboard
90 Qwen2-7B-Instruct 3.25% Jun 2024 huggingface.co Official leaderboard
91 Gemma 3 1B 0.83% Mar 2025 huggingface.co Official leaderboard
#ModelScoreSettingDateLinkSource
1 GPT-5.5 98% high Apr 2026 developers.openai.com Official leaderboard
2 Claude Fable 5 97.8% high Jun 2026 anthropic.com Official leaderboard
3 Claude Opus 4.8 97.4% high May 2026 anthropic.com Official leaderboard
4 Gemini 3.7 Flash 97.3% high Aug 2026 ai.google.dev Official leaderboard
5 Gemini 3 Pro 97.3% high Nov 2025 ai.google.dev Official leaderboard
6 Claude Opus 4.7 97.1% high Apr 2026 anthropic.com Official leaderboard
7 GPT-5.6 Sol 97% high Jul 2026 developers.openai.com Official leaderboard
8 Gemini 3.5 Flash 96.9% high May 2026 ai.google.dev Official leaderboard
9 Gemini 3.1 Pro 96.7% high Feb 2026 ai.google.dev Official leaderboard
10 GPT-5.4 96.7% high Mar 2026 huggingface.co Official leaderboard
11 Gemini 3.6 Flash 96.5% high Jul 2026 ai.google.dev Official leaderboard
12 Gemini 3 Flash 95.7% high Dec 2025 ai.google.dev Official leaderboard
13 GPT-5 95.2% high Aug 2025 huggingface.co Official leaderboard
14 GPT-5.1 94.8% high Nov 2025 huggingface.co Official leaderboard
15 GPT-5.6 Luna 94.5% high Jul 2026 developers.openai.com Official leaderboard
16 Gemini 3.1 Flash-Lite 92.7% high Mar 2026 ai.google.dev Official leaderboard
17 GPT-5.6 Terra 92.5% high Jul 2026 developers.openai.com Official leaderboard
18 DeepSeek-V4-Pro 92.2% on Apr 2026 huggingface.co Official leaderboard
19 Claude Opus 4.6 92.1% high Feb 2026 anthropic.com Official leaderboard
20 GPT-5.2 91.3% high Dec 2025 huggingface.co Official leaderboard
21 Gemini 2.5 Pro Jun 2025 90.8% on Jun 2025 huggingface.co Official leaderboard
22 Gemini 2.0 Pro 89.7% on Feb 2025 huggingface.co Official leaderboard
23 Claude Opus 4.5 89.7% high Nov 2025 huggingface.co Official leaderboard
24 Claude 3.5 Sonnet 89.2% Jun 2024 huggingface.co Official leaderboard
25 Claude 3.7 Sonnet 88.3% Feb 2025 huggingface.co Official leaderboard
26 GPT-4o May 2024 87.8% May 2024 huggingface.co Official leaderboard
27 GPT-4o Aug 2024 87.8% Aug 2024 huggingface.co Official leaderboard
28 Gemini 2.0 Flash Feb 2025 86.6% Feb 2025 huggingface.co Official leaderboard
29 GPT-4.1 84.6% Apr 2025 huggingface.co Official leaderboard
30 o1-preview 84.4% on Sep 2024 huggingface.co Official leaderboard
31 Claude 3 Opus 84.2% Feb 2024 huggingface.co Official leaderboard
32 Gemini 2.5 Flash Sep 2025 82.5% high Sep 2025 huggingface.co Official leaderboard
33 Grok 4.3 82.4% on Apr 2026 docs.x.ai Official leaderboard
34 Kimi K2.5 81.2% on Jan 2026 kimi.com Official leaderboard
35 Claude Sonnet 4.6 80.4% high Feb 2026 anthropic.com Official leaderboard
36 GLM-5.2 80% on Jun 2026 huggingface.co Official leaderboard
37 Claude Sonnet 4.5 79.5% high Sep 2025 huggingface.co Official leaderboard
38 GPT-5 mini 77.7% high Aug 2025 huggingface.co Official leaderboard
39 DeepSeek-V3 Mar 2025 77.3% Mar 2025 huggingface.co Official leaderboard
40 GPT-4 Turbo Apr 2024 76.6% Apr 2024 huggingface.co Official leaderboard
41 Gemini 2.5 Flash-Lite Sep 2025 74.1% high Sep 2025 huggingface.co Official leaderboard
42 Claude Sonnet 4 74.1% on May 2025 huggingface.co Official leaderboard
43 Mistral Large 3 73.4% Dec 2025 huggingface.co Official leaderboard
44 o4-mini 73.1% on Apr 2025 huggingface.co Official leaderboard
45 Claude 3.5 Haiku 70.6% Oct 2024 huggingface.co Official leaderboard
46 GPT-4.1 mini 69.6% Apr 2025 huggingface.co Official leaderboard
47 DeepSeek-R1 May 2025 68.5% on May 2025 huggingface.co Official leaderboard
48 Gemma 4 31B IT 64.8% on Apr 2026 huggingface.co Official leaderboard
49 GPT-4o mini 63.9% Jul 2024 huggingface.co Official leaderboard
50 MiniMax-M2.5 59.6% on Feb 2026 minimax.io Official leaderboard
51 Claude 2.1 55.2% Nov 2023 huggingface.co Official leaderboard
52 GPT-4 Mar 2023 54.6% Mar 2023 huggingface.co Official leaderboard
53 Gemma 4 26B A4B 53.6% on Apr 2026 huggingface.co Official leaderboard
54 Mistral Medium 3.1 53.1% Aug 2025 huggingface.co Official leaderboard
55 GPT-4 Jun 2023 50.7% Jun 2023 huggingface.co Official leaderboard
56 Gemma 3 27B 50.4% Mar 2025 huggingface.co Official leaderboard
57 Llama 3.3 70B Instruct 48.2% Dec 2024 huggingface.co Official leaderboard
58 GPT-5 nano 47.7% high Aug 2025 huggingface.co Official leaderboard
59 Llama 3.1-70B 46.3% Jul 2024 huggingface.co Official leaderboard
60 o3-mini 46% on Jan 2025 huggingface.co Official leaderboard
61 Llama-3.1-Nemotron-70B-Instruct 46% Oct 2024 huggingface.co Official leaderboard
62 GPT-4.1 nano 39.9% Apr 2025 huggingface.co Official leaderboard
63 GPT-3.5 Turbo Jan 2024 39.1% Jan 2024 huggingface.co Official leaderboard
64 Qwen3.5 397B-A17B 38.2% on Feb 2026 huggingface.co Official leaderboard
65 o1-mini 36.5% on Sep 2024 huggingface.co Official leaderboard
66 Gemma 3 12B 32.9% Mar 2025 huggingface.co Official leaderboard
67 Mistral NeMo 27.2% Jul 2024 huggingface.co Official leaderboard
68 Llama 3-70B 23.8% Apr 2024 huggingface.co Official leaderboard
69 DeepSeek-R1-Distill-Llama-70B 21.5% on Jan 2025 huggingface.co Official leaderboard
70 Llama 3.1-8B 18.8% Jul 2024 huggingface.co Official leaderboard
71 Ministral 3B 18% Oct 2024 huggingface.co Official leaderboard
72 Gemma 3 4B 17.1% Mar 2025 huggingface.co Official leaderboard
73 Gemma 2 9B 11% Jun 2024 huggingface.co Official leaderboard
74 Llama 3.2 3B 10.2% Sep 2024 huggingface.co Official leaderboard
75 Llama 3.2 1B 9.71% Sep 2024 huggingface.co Official leaderboard
76 Llama 3-8B 7.17% Apr 2024 huggingface.co Official leaderboard
77 Qwen2-72B 4.46% Jun 2024 huggingface.co Official leaderboard
78 Qwen2-7B-Instruct 3.25% Jun 2024 huggingface.co Official leaderboard
79 Gemma 3 1B 0.83% Mar 2025 huggingface.co Official leaderboard

10 models have held number one on the official leaderboard since its release in November 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 May 2024, by GPT-4o
GPT-5.5 98%

Source: Official Miðeind Icelandic Case Declension leaderboard

GPT-4o on the official leaderboard passed 85% in May 2024

That was 6 months after its release. By the same age, 9% 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%) GPT-5.5 GPT-5 Gemini 2.5 Pro Claude 3.5 Sonnet Claude 3 Opus Claude 2.1 GPT-4 0% 20% 40% 60% 80% 100% 2023 2024 2025 2026 Near the ceiling (85%) GPT-5.5 Gemini 2.5 Pro

Source: Official Miðeind Icelandic Case Declension 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 (91 entries).
  • Pass Models scored 79 models we track, plus 12 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 98%, past 85% since May 2024.
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 12 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