Benchmark Almanac

Miðeind Icelandic Grammar Correction

Claude Fable 5.2 83.9% leads since November 2026
Not in model cards

Models correct an Icelandic sentence or return it unchanged. Half the sentences contain one error. The score measures exact agreement with the expected sentence.

Passes test 2. Miðeind reports scores for 60 named models on its current leaderboard. The board reports separate scores for sentences with and without errors.

Made by
Miðeind
Released
September 2026
Kind
Chat and instructions in many languages
Level
School
Scored
Checked automatically
Answers
Open-ended
Tasks
Kept private
Measures
Capability
Language
Icelandic
In and out
Text → Text
Interaction
Single turn
Items from
Written by experts
Refresh
One fixed set
Run by
A host
Strongest model tested
Claude Fable 5.2 #2 on the capability index
Last model added
November 2026
Scores refreshed
November 20th, 2026
62
Entries on its own leaderboard, 61 models we track

Claude Fable 5.2 leads with 83.9% Claude Fable 5.2 leads with 83.9%

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. 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. This view shows only models we track, each at its best.

  1. 1 Claude Fable 5.2 high 83.9%
  2. 2 Claude Opus 5 high 83%
  3. 3 Claude Fable 5.1 high 81.7%
  4. 4 GPT-6 Astra high 81.2%
  5. 5 Claude Opus 5.5 high 80.7%
  6. 6 Gemini 4 Argon high 80.7%
  7. 7 Gemini 4 Neon low 80.7%
  8. 8 GPT-6.1 Terra high 79.4%
  9. 9 GPT-6.2 Sol high 78.4%
  10. 10 Gemini 4 Hydrogen high 78%
  11. 11 GPT-6 Sol high 74.8%
  12. 12 GPT-6.1 Sol high 74.3%
  13. 13 Gemini 3.1 Pro low 73.4%
  14. 14 Claude Opus 4.8 high 72.9%
  15. 15 Gemini 4 Carbon low 71.1%
  16. 16 GPT-5.6 Sol high 70.6%
  17. 17 Claude Sonnet 5.5 high 67.4%
  18. 18 Gemini 3.5 Flash high 67%
  19. 19 Gemini 3.7 Flash high 66.5%
  20. 20 Claude Haiku 5.5 high 64.2%
  21. 21 Gemini 3.8 Flash high 63.3%
  22. 22 Claude Sonnet 5 off 59.6%
  23. 23 GPT-5.6 Terra high 58.3%
  24. 24 GPT-5.6 Luna high 57.8%
  25. 25 Inkling 2 Small high 56.9%
  26. 26 Grok 4.7 high 56.4%
  27. 27 GPT-6 Luna high 56%
  28. 28 DeepSeek V4.1 Flash on 54.6%
  29. 29 DeepSeek-V4-Pro off 51.4%
  30. 30 Bielik 11B v3.1 Instruct off 49.5%
  31. 31 Gemini 3.5 Flash-Lite off 49.1%
  32. 32 Muse Spark Open high 49.1%
  33. 33 Kimi K3 off 46.3%
  34. 34 Qwen3.8-2.4T-A95B on 45%
  35. 35 MiniMax-M3.1-Flash off 44.5%
  36. 36 Gemma 4 E2B off 44%
  37. 37 Gemini 2.5 Flash-Lite Sep 2025, off 42.7%
  38. 38 Gemini 2.5 Pro Jun 2025, high 41.3%
  39. 39 Inkling off 40.8%
  40. 40 MiniMax-M3 on 37.2%
  41. 41 Gemma 4 12B off 36.7%
  42. 42 Gemma 4 26B A4B off 36.2%
  43. 43 GPT-5.2 off 35.8%
  44. 44 GLM-5.3-Flash on 35.3%
  45. 45 GPT-4.1 nano none 34.4%
  46. 46 GPT-4o Mar 2025, none 33.9%
  47. 47 Mistral Small 4 off 33.9%
  48. 48 Apertus 1.5 70B on 33.5%
  49. 49 Inkling-Small high 30.7%
  50. 50 Gemma 4 E4B off 29.8%
  51. 51 Qwen 3.8 27B off 27.1%
  52. 52 Mistral Large 4 low 25.7%
  53. 53 Apertus 1.5 8B on 21.6%
  54. 54 Mistral Medium 3.5 off 20.2%
  55. 55 Gemma 4 31B IT on 19.7%
  56. 56 Mistral Large 3 none 18.8%
  57. 57 GPT-3.5 Turbo Jan 2024, none 17.4%
  58. 58 Llama 3.1-70B none 17%
  59. 59 Claude Haiku 4.5 off 14.7%
  60. 60 Ripple-65B Preview on 10.6%
  61. 61 Llama 3.1-8B none 8.26%
  62. 62 Apertus 8B Instruct none 2.29%

Source: Official Miðeind Icelandic Grammar Correction leaderboard

  1. 1 Claude Fable 5.2 high 83.9%
  2. 2 Claude Opus 5 high 83%
  3. 3 Claude Fable 5.1 high 81.7%
  4. 4 GPT-6 Astra high 81.2%
  5. 5 Claude Opus 5.5 high 80.7%
  6. 6 Gemini 4 Argon high 80.7%
  7. 7 Gemini 4 Neon low 80.7%
  8. 8 GPT-6.1 Terra high 79.4%
  9. 9 GPT-6.2 Sol high 78.4%
  10. 10 Gemini 4 Hydrogen high 78%
  11. 11 GPT-6 Sol high 74.8%
  12. 12 GPT-6.1 Sol high 74.3%
  13. 13 Gemini 3.1 Pro low 73.4%
  14. 14 Claude Opus 4.8 high 72.9%
  15. 15 Gemini 4 Carbon low 71.1%
  16. 16 GPT-5.6 Sol high 70.6%
  17. 17 Claude Sonnet 5.5 high 67.4%
  18. 18 Gemini 3.5 Flash high 67%
  19. 19 Gemini 3.7 Flash high 66.5%
  20. 20 Claude Haiku 5.5 high 64.2%
  21. 21 Gemini 3.8 Flash high 63.3%
  22. 22 Claude Sonnet 5 off 59.6%
  23. 23 GPT-5.6 Terra high 58.3%
  24. 24 GPT-5.6 Luna high 57.8%
  25. 25 Inkling 2 Small high 56.9%
  26. 26 Grok 4.7 high 56.4%
  27. 27 GPT-6 Luna high 56%
  28. 28 DeepSeek V4.1 Flash on 54.6%
  29. 29 DeepSeek-V4-Pro off 51.4%
  30. 30 Gemini 3.5 Flash-Lite off 49.1%
  31. 31 Muse Spark Open high 49.1%
  32. 32 Kimi K3 off 46.3%
  33. 33 Qwen3.8-2.4T-A95B on 45%
  34. 34 MiniMax-M3.1-Flash off 44.5%
  35. 35 Gemma 4 E2B off 44%
  36. 36 Gemini 2.5 Flash-Lite Sep 2025, off 42.7%
  37. 37 Gemini 2.5 Pro Jun 2025, high 41.3%
  38. 38 Inkling off 40.8%
  39. 39 MiniMax-M3 on 37.2%
  40. 40 Gemma 4 12B off 36.7%
  41. 41 Gemma 4 26B A4B off 36.2%
  42. 42 GPT-5.2 off 35.8%
  43. 43 GLM-5.3-Flash on 35.3%
  44. 44 GPT-4.1 nano none 34.4%
  45. 45 GPT-4o Mar 2025, none 33.9%
  46. 46 Mistral Small 4 off 33.9%
  47. 47 Apertus 1.5 70B on 33.5%
  48. 48 Inkling-Small high 30.7%
  49. 49 Gemma 4 E4B off 29.8%
  50. 50 Qwen 3.8 27B off 27.1%
  51. 51 Mistral Large 4 low 25.7%
  52. 52 Apertus 1.5 8B on 21.6%
  53. 53 Mistral Medium 3.5 off 20.2%
  54. 54 Gemma 4 31B IT on 19.7%
  55. 55 Mistral Large 3 none 18.8%
  56. 56 GPT-3.5 Turbo Jan 2024, none 17.4%
  57. 57 Llama 3.1-70B none 17%
  58. 58 Claude Haiku 4.5 off 14.7%
  59. 59 Ripple-65B Preview on 10.6%
  60. 60 Llama 3.1-8B none 8.26%
  61. 61 Apertus 8B Instruct none 2.29%

Source: Official Miðeind Icelandic Grammar Correction leaderboard

#EntryScoreSettingReleasedRun onLinkSource
1 Claude Fable 5.2 83.9% high Nov 2026 Nov 2026 anthropic.com Official leaderboard
2 Claude Opus 5 83% high Jul 2026 Sep 2026 anthropic.com Official leaderboard
3 Claude Fable 5.1 81.7% high Sep 2026 Sep 2026 anthropic.com Official leaderboard
4 GPT-6 Astra 81.2% high Sep 2026 Sep 2026 developers.openai.com Official leaderboard
5 Claude Opus 5.5 80.7% high Sep 2026 Sep 2026 anthropic.com Official leaderboard
6 Gemini 4 Argon 80.7% high Sep 2026 Oct 2026 ai.google.dev Official leaderboard
7 Gemini 4 Neon 80.7% low Nov 2026 Nov 2026 ai.google.dev Official leaderboard
8 GPT-6.1 Terra 79.4% high Oct 2026 Oct 2026 developers.openai.com Official leaderboard
9 GPT-6.2 Sol 78.4% high Nov 2026 Nov 2026 developers.openai.com Official leaderboard
10 Gemini 4 Hydrogen 78% high Nov 2026 Nov 2026 ai.google.dev Official leaderboard
11 GPT-6 Sol 74.8% high Sep 2026 Sep 2026 developers.openai.com Official leaderboard
12 GPT-6.1 Sol 74.3% high Sep 2026 Sep 2026 developers.openai.com Official leaderboard
13 Gemini 3.1 Pro 73.4% low Feb 2026 Sep 2026 ai.google.dev Official leaderboard
14 Claude Opus 4.8 72.9% high May 2026 Sep 2026 anthropic.com Official leaderboard
15 Gemini 4 Carbon 71.1% low Oct 2026 Nov 2026 ai.google.dev Official leaderboard
16 GPT-5.6 Sol 70.6% high Jul 2026 Sep 2026 developers.openai.com Official leaderboard
17 Claude Sonnet 5.5 67.4% high Sep 2026 Sep 2026 anthropic.com Official leaderboard
18 Gemini 3.5 Flash 67% high May 2026 Sep 2026 ai.google.dev Official leaderboard
19 Gemini 3.7 Flash 66.5% high Aug 2026 Sep 2026 ai.google.dev Official leaderboard
20 Claude Haiku 5.5 64.2% high Oct 2026 Oct 2026 anthropic.com Official leaderboard
21 Gemini 3.8 Flash 63.3% high Sep 2026 Sep 2026 ai.google.dev Official leaderboard
22 Claude Sonnet 5 59.6% off Jun 2026 Sep 2026 anthropic.com Official leaderboard
23 GPT-5.6 Terra 58.3% high Jul 2026 Sep 2026 developers.openai.com Official leaderboard
24 GPT-5.6 Luna 57.8% high Jul 2026 Sep 2026 developers.openai.com Official leaderboard
25 Inkling 2 Small 56.9% high Nov 2026 Nov 2026 huggingface.co Official leaderboard
26 Grok 4.7 56.4% high Sep 2026 Nov 2026 docs.x.ai Official leaderboard
27 GPT-6 Luna 56% high Sep 2026 Sep 2026 developers.openai.com Official leaderboard
28 DeepSeek V4.1 Flash 54.6% on Sep 2026 Oct 2026 huggingface.co Official leaderboard
29 DeepSeek-V4-Pro 51.4% off Apr 2026 Sep 2026 huggingface.co Official leaderboard
30 Bielik 11B v3.1 Instruct 11.2B 49.5% off Nov 2026 huggingface.co Official leaderboard
31 Gemini 3.5 Flash-Lite 49.1% off Jul 2026 Sep 2026 ai.google.dev Official leaderboard
32 Muse Spark Open 49.1% high Oct 2026 Oct 2026 huggingface.co Official leaderboard
33 Kimi K3 46.3% off Jul 2026 Sep 2026 huggingface.co Official leaderboard
34 Qwen3.8-2.4T-A95B 45% on Aug 2026 Sep 2026 huggingface.co Official leaderboard
35 MiniMax-M3.1-Flash 44.5% off Oct 2026 Oct 2026 huggingface.co Official leaderboard
36 Gemma 4 E2B 44% off Mar 2026 Oct 2026 huggingface.co Official leaderboard
37 Gemini 2.5 Flash-Lite Sep 2025 42.7% off Sep 2025 Sep 2026 ai.google.dev Official leaderboard
38 Gemini 2.5 Pro Jun 2025 41.3% high Jun 2025 Sep 2026 ai.google.dev Official leaderboard
39 Inkling 40.8% off Jul 2026 Sep 2026 huggingface.co Official leaderboard
40 MiniMax-M3 37.2% on Jun 2026 Sep 2026 huggingface.co Official leaderboard
41 Gemma 4 12B 36.7% off Jun 2026 Nov 2026 huggingface.co Official leaderboard
42 Gemma 4 26B A4B 36.2% off Apr 2026 Oct 2026 huggingface.co Official leaderboard
43 GPT-5.2 35.8% off Dec 2025 Sep 2026 platform.openai.com Official leaderboard
44 GLM-5.3-Flash 35.3% on Aug 2026 Sep 2026 huggingface.co Official leaderboard
45 GPT-4.1 nano 34.4% none Apr 2025 Sep 2026 platform.openai.com Official leaderboard
46 GPT-4o Mar 2025 33.9% none Mar 2025 Sep 2026 platform.openai.com Official leaderboard
47 Mistral Small 4 33.9% off Mar 2026 Sep 2026 mistral.ai Official leaderboard
48 Apertus 1.5 70B 33.5% on Jul 2026 Sep 2026 huggingface.co Official leaderboard
49 Inkling-Small 30.7% high Jul 2026 Sep 2026 huggingface.co Official leaderboard
50 Gemma 4 E4B 29.8% off Mar 2026 Oct 2026 huggingface.co Official leaderboard
51 Qwen 3.8 27B 27.1% off Aug 2026 Oct 2026 huggingface.co Official leaderboard
52 Mistral Large 4 25.7% low Oct 2026 Oct 2026 mistral.ai Official leaderboard
53 Apertus 1.5 8B 21.6% on Jul 2026 Sep 2026 huggingface.co Official leaderboard
54 Mistral Medium 3.5 20.2% off Apr 2026 Sep 2026 mistral.ai Official leaderboard
55 Gemma 4 31B IT 19.7% on Apr 2026 Sep 2026 deepmind.google Official leaderboard
56 Mistral Large 3 18.8% none Dec 2025 Sep 2026 mistral.ai Official leaderboard
57 GPT-3.5 Turbo Jan 2024 17.4% none Jan 2024 Sep 2026 platform.openai.com Official leaderboard
58 Llama 3.1-70B 17% none Jul 2024 Oct 2026 huggingface.co Official leaderboard
59 Claude Haiku 4.5 14.7% off Oct 2025 Sep 2026 anthropic.com Official leaderboard
60 Ripple-65B Preview 10.6% on Nov 2026 Nov 2026 huggingface.co Official leaderboard
61 Llama 3.1-8B 8.26% none Jul 2024 Oct 2026 huggingface.co Official leaderboard
62 Apertus 8B Instruct 2.29% none Sep 2025 Sep 2026 huggingface.co Official leaderboard
#ModelScoreSettingReleasedRun onLinkSource
1 Claude Fable 5.2 83.9% high Nov 2026 Nov 2026 anthropic.com Official leaderboard
2 Claude Opus 5 83% high Jul 2026 Sep 2026 anthropic.com Official leaderboard
3 Claude Fable 5.1 81.7% high Sep 2026 Sep 2026 anthropic.com Official leaderboard
4 GPT-6 Astra 81.2% high Sep 2026 Sep 2026 developers.openai.com Official leaderboard
5 Claude Opus 5.5 80.7% high Sep 2026 Sep 2026 anthropic.com Official leaderboard
6 Gemini 4 Argon 80.7% high Sep 2026 Oct 2026 ai.google.dev Official leaderboard
7 Gemini 4 Neon 80.7% low Nov 2026 Nov 2026 ai.google.dev Official leaderboard
8 GPT-6.1 Terra 79.4% high Oct 2026 Oct 2026 developers.openai.com Official leaderboard
9 GPT-6.2 Sol 78.4% high Nov 2026 Nov 2026 developers.openai.com Official leaderboard
10 Gemini 4 Hydrogen 78% high Nov 2026 Nov 2026 ai.google.dev Official leaderboard
11 GPT-6 Sol 74.8% high Sep 2026 Sep 2026 developers.openai.com Official leaderboard
12 GPT-6.1 Sol 74.3% high Sep 2026 Sep 2026 developers.openai.com Official leaderboard
13 Gemini 3.1 Pro 73.4% low Feb 2026 Sep 2026 ai.google.dev Official leaderboard
14 Claude Opus 4.8 72.9% high May 2026 Sep 2026 anthropic.com Official leaderboard
15 Gemini 4 Carbon 71.1% low Oct 2026 Nov 2026 ai.google.dev Official leaderboard
16 GPT-5.6 Sol 70.6% high Jul 2026 Sep 2026 developers.openai.com Official leaderboard
17 Claude Sonnet 5.5 67.4% high Sep 2026 Sep 2026 anthropic.com Official leaderboard
18 Gemini 3.5 Flash 67% high May 2026 Sep 2026 ai.google.dev Official leaderboard
19 Gemini 3.7 Flash 66.5% high Aug 2026 Sep 2026 ai.google.dev Official leaderboard
20 Claude Haiku 5.5 64.2% high Oct 2026 Oct 2026 anthropic.com Official leaderboard
21 Gemini 3.8 Flash 63.3% high Sep 2026 Sep 2026 ai.google.dev Official leaderboard
22 Claude Sonnet 5 59.6% off Jun 2026 Sep 2026 anthropic.com Official leaderboard
23 GPT-5.6 Terra 58.3% high Jul 2026 Sep 2026 developers.openai.com Official leaderboard
24 GPT-5.6 Luna 57.8% high Jul 2026 Sep 2026 developers.openai.com Official leaderboard
25 Inkling 2 Small 56.9% high Nov 2026 Nov 2026 huggingface.co Official leaderboard
26 Grok 4.7 56.4% high Sep 2026 Nov 2026 docs.x.ai Official leaderboard
27 GPT-6 Luna 56% high Sep 2026 Sep 2026 developers.openai.com Official leaderboard
28 DeepSeek V4.1 Flash 54.6% on Sep 2026 Oct 2026 huggingface.co Official leaderboard
29 DeepSeek-V4-Pro 51.4% off Apr 2026 Sep 2026 huggingface.co Official leaderboard
30 Gemini 3.5 Flash-Lite 49.1% off Jul 2026 Sep 2026 ai.google.dev Official leaderboard
31 Muse Spark Open 49.1% high Oct 2026 Oct 2026 huggingface.co Official leaderboard
32 Kimi K3 46.3% off Jul 2026 Sep 2026 huggingface.co Official leaderboard
33 Qwen3.8-2.4T-A95B 45% on Aug 2026 Sep 2026 huggingface.co Official leaderboard
34 MiniMax-M3.1-Flash 44.5% off Oct 2026 Oct 2026 huggingface.co Official leaderboard
35 Gemma 4 E2B 44% off Mar 2026 Oct 2026 huggingface.co Official leaderboard
36 Gemini 2.5 Flash-Lite Sep 2025 42.7% off Sep 2025 Sep 2026 ai.google.dev Official leaderboard
37 Gemini 2.5 Pro Jun 2025 41.3% high Jun 2025 Sep 2026 ai.google.dev Official leaderboard
38 Inkling 40.8% off Jul 2026 Sep 2026 huggingface.co Official leaderboard
39 MiniMax-M3 37.2% on Jun 2026 Sep 2026 huggingface.co Official leaderboard
40 Gemma 4 12B 36.7% off Jun 2026 Nov 2026 huggingface.co Official leaderboard
41 Gemma 4 26B A4B 36.2% off Apr 2026 Oct 2026 huggingface.co Official leaderboard
42 GPT-5.2 35.8% off Dec 2025 Sep 2026 platform.openai.com Official leaderboard
43 GLM-5.3-Flash 35.3% on Aug 2026 Sep 2026 huggingface.co Official leaderboard
44 GPT-4.1 nano 34.4% none Apr 2025 Sep 2026 platform.openai.com Official leaderboard
45 GPT-4o Mar 2025 33.9% none Mar 2025 Sep 2026 platform.openai.com Official leaderboard
46 Mistral Small 4 33.9% off Mar 2026 Sep 2026 mistral.ai Official leaderboard
47 Apertus 1.5 70B 33.5% on Jul 2026 Sep 2026 huggingface.co Official leaderboard
48 Inkling-Small 30.7% high Jul 2026 Sep 2026 huggingface.co Official leaderboard
49 Gemma 4 E4B 29.8% off Mar 2026 Oct 2026 huggingface.co Official leaderboard
50 Qwen 3.8 27B 27.1% off Aug 2026 Oct 2026 huggingface.co Official leaderboard
51 Mistral Large 4 25.7% low Oct 2026 Oct 2026 mistral.ai Official leaderboard
52 Apertus 1.5 8B 21.6% on Jul 2026 Sep 2026 huggingface.co Official leaderboard
53 Mistral Medium 3.5 20.2% off Apr 2026 Sep 2026 mistral.ai Official leaderboard
54 Gemma 4 31B IT 19.7% on Apr 2026 Sep 2026 deepmind.google Official leaderboard
55 Mistral Large 3 18.8% none Dec 2025 Sep 2026 mistral.ai Official leaderboard
56 GPT-3.5 Turbo Jan 2024 17.4% none Jan 2024 Sep 2026 platform.openai.com Official leaderboard
57 Llama 3.1-70B 17% none Jul 2024 Oct 2026 huggingface.co Official leaderboard
58 Claude Haiku 4.5 14.7% off Oct 2025 Sep 2026 anthropic.com Official leaderboard
59 Ripple-65B Preview 10.6% on Nov 2026 Nov 2026 huggingface.co Official leaderboard
60 Llama 3.1-8B 8.26% none Jul 2024 Oct 2026 huggingface.co Official leaderboard
61 Apertus 8B Instruct 2.29% none Sep 2025 Sep 2026 huggingface.co Official leaderboard

Rankings by part

We read 2 parts from the benchmark’s own leaderboard: With error and Error free. Each ranks models on one part of the test. Gemini 3.1 Pro leads With error and Claude Fable 5.1 leads Error free. Arrows show how far a model moved from its overall rank among the models we track.

With error

61 models we track have a score here.

  1. 1 Gemini 3.1 Pro ↑12 79.8%
  2. 2 Claude Opus 5 79.8%
  3. 3 Claude Fable 5.2 ↓2 79.8%
  4. 4 GPT-6 Astra 78.9%
  5. 5 GPT-6.2 Sol ↑4 78%
  6. 6 Gemini 3.8 Flash ↑15 77.1%
  7. 7 Gemini 4 Argon ↓1 77.1%
  8. 8 Claude Opus 5.5 ↓3 75.2%
  9. 9 GPT-6.1 Sol ↑3 75.2%
  10. 10 Gemini 4 Carbon ↑5 75.2%

More effort adds a median 5 points

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
Claude Fable 5.2
low high
83.9%
Claude Opus 5
off high
83%
GPT-6 Astra
low
81.2%
Gemini 4 Argon
off high
80.7%
GPT-6.1 Terra
off high
79.4%
GPT-6.2 Sol
low high
78.4%
Gemini 4 Hydrogen
low high
78%
GPT-6 Sol
off high
74.8%
GPT-6.1 Sol
low
74.3%
Gemini 3.1 Pro
low high
73.4%
Claude Opus 4.8
off
72.9%
GPT-5.6 Sol
off high
70.6%

Source: Official Miðeind Icelandic Grammar Correction leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude Fable 5.2 high 83.9%
Claude Opus 5 high 83%
Claude Fable 5.1 high 81.7%
GPT-6 Astra high 81.2%
Claude Opus 5.5 high 80.7%
Gemini 4 Argon high 80.7%
Gemini 4 Neon low 80.7%
GPT-6.1 Terra high 79.4%
Claude Fable 5.2 low 78.9%
GPT-6 Astra low 78.4%
GPT-6.2 Sol high 78.4%
Gemini 4 Hydrogen high 78%
Claude Opus 5 off 75.7%
GPT-6 Sol high 74.8%
Gemini 4 Argon off 74.8%
GPT-6.1 Sol high 74.3%
Gemini 3.1 Pro low 73.4%
Claude Opus 4.8 high 72.9%
Gemini 4 Hydrogen low 72.9%
Claude Opus 4.8 off 72.5%
GPT-6.1 Sol low 72%
Gemini 4 Carbon low 71.1%
GPT-5.6 Sol high 70.6%
GPT-6.2 Sol low 70.2%
Gemini 3.1 Pro high 68.8%
GPT-6 Sol off 68.3%
Claude Sonnet 5.5 high 67.4%
Gemini 3.5 Flash high 67%
Gemini 3.7 Flash high 66.5%
Claude Haiku 5.5 high 64.2%
GPT-6.1 Terra off 63.8%
Gemini 3.8 Flash high 63.3%
Gemini 3.8 Flash low 61.9%
Gemini 3.7 Flash low 61.5%
Claude Sonnet 5 off 59.6%
Claude Sonnet 5.5 off 59.6%
GPT-5.6 Terra high 58.3%
Claude Sonnet 5 high 57.8%
GPT-5.6 Luna high 57.8%
Claude Haiku 5.5 off 57.3%
Inkling 2 Small high 56.9%
Grok 4.7 high 56.4%
GPT-5.6 Sol off 56%
GPT-6 Luna high 56%
DeepSeek V4.1 Flash on 54.6%
Grok 4.7 low 54.1%
Inkling 2 Small minimal 52.3%
Gemini 3.5 Flash off 51.8%
DeepSeek-V4-Pro off 51.4%
GPT-6 Luna off 51.4%
GPT-5.6 Luna off 49.1%
Gemini 3.5 Flash-Lite off 49.1%
Muse Spark Open high 49.1%
GPT-5.6 Terra off 48.6%
DeepSeek V4.1 Flash off 48.6%
Kimi K3 off 46.3%
Gemini 3.5 Flash-Lite high 45%
Qwen3.8-2.4T-A95B on 45%
MiniMax-M3.1-Flash off 44.5%
Gemma 4 E2B off 44%
Kimi K3 on 44%
DeepSeek-V4-Pro on 43.6%
Gemini 2.5 Flash-Lite Sep 2025 off 42.7%
Gemini 2.5 Pro Jun 2025 high 41.3%
Inkling off 40.8%
Muse Spark Open off 40.8%
Inkling high 40.4%
MiniMax-M3 on 37.2%
Gemini 2.5 Pro Jun 2025 minimal 36.7%
Gemma 4 12B off 36.7%
Gemma 4 26B A4B off 36.2%
GPT-5.2 off 35.8%
GPT-5.2 high 35.3%
GLM-5.3-Flash on 35.3%
GPT-4.1 nano none 34.4%
GPT-4o Mar 2025 none 33.9%
Mistral Small 4 off 33.9%
Apertus 1.5 70B on 33.5%
Inkling-Small high 30.7%
Gemma 4 E2B on 29.8%
Gemma 4 E4B off 29.8%
Qwen 3.8 27B off 27.1%
Mistral Large 4 low 25.7%
Gemma 4 E4B on 25.2%
Mistral Large 4 high 25.2%
MiniMax-M3.1-Flash high 24.8%
Apertus 1.5 8B on 21.6%
Mistral Medium 3.5 off 20.2%
Gemini 2.5 Flash-Lite Sep 2025 high 19.7%
Gemma 4 31B IT on 19.7%
Qwen 3.8 27B xhigh 19.3%
Mistral Large 3 none 18.8%
GPT-3.5 Turbo Jan 2024 none 17.4%
Gemma 4 12B on 17.4%
Llama 3.1-70B none 17%
Gemma 4 26B A4B on 16.5%
Mistral Small 4 high 15.6%
Claude Haiku 4.5 off 14.7%
Claude Haiku 4.5 high 11.5%
Mistral Medium 3.5 high 10.6%
Ripple-65B Preview on 10.6%
Llama 3.1-8B none 8.26%
Apertus 8B Instruct none 2.29%

More effort adds a median 5 points

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
Claude Fable 5.2
low high
83.9%
Claude Opus 5
off high
83%
GPT-6 Astra
low
81.2%
Gemini 4 Argon
off high
80.7%
GPT-6.1 Terra
off high
79.4%
GPT-6.2 Sol
low high
78.4%
Gemini 4 Hydrogen
low high
78%
GPT-6 Sol
off high
74.8%
GPT-6.1 Sol
low
74.3%
Gemini 3.1 Pro
low high
73.4%
Claude Opus 4.8
off
72.9%
GPT-5.6 Sol
off high
70.6%

Source: Official Miðeind Icelandic Grammar Correction leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude Fable 5.2 high 83.9%
Claude Opus 5 high 83%
Claude Fable 5.1 high 81.7%
GPT-6 Astra high 81.2%
Claude Opus 5.5 high 80.7%
Gemini 4 Argon high 80.7%
Gemini 4 Neon low 80.7%
GPT-6.1 Terra high 79.4%
Claude Fable 5.2 low 78.9%
GPT-6 Astra low 78.4%
GPT-6.2 Sol high 78.4%
Gemini 4 Hydrogen high 78%
Claude Opus 5 off 75.7%
GPT-6 Sol high 74.8%
Gemini 4 Argon off 74.8%
GPT-6.1 Sol high 74.3%
Gemini 3.1 Pro low 73.4%
Claude Opus 4.8 high 72.9%
Gemini 4 Hydrogen low 72.9%
Claude Opus 4.8 off 72.5%
GPT-6.1 Sol low 72%
Gemini 4 Carbon low 71.1%
GPT-5.6 Sol high 70.6%
GPT-6.2 Sol low 70.2%
Gemini 3.1 Pro high 68.8%
GPT-6 Sol off 68.3%
Claude Sonnet 5.5 high 67.4%
Gemini 3.5 Flash high 67%
Gemini 3.7 Flash high 66.5%
Claude Haiku 5.5 high 64.2%
GPT-6.1 Terra off 63.8%
Gemini 3.8 Flash high 63.3%
Gemini 3.8 Flash low 61.9%
Gemini 3.7 Flash low 61.5%
Claude Sonnet 5 off 59.6%
Claude Sonnet 5.5 off 59.6%
GPT-5.6 Terra high 58.3%
Claude Sonnet 5 high 57.8%
GPT-5.6 Luna high 57.8%
Claude Haiku 5.5 off 57.3%
Inkling 2 Small high 56.9%
Grok 4.7 high 56.4%
GPT-5.6 Sol off 56%
GPT-6 Luna high 56%
DeepSeek V4.1 Flash on 54.6%
Grok 4.7 low 54.1%
Inkling 2 Small minimal 52.3%
Gemini 3.5 Flash off 51.8%
DeepSeek-V4-Pro off 51.4%
GPT-6 Luna off 51.4%
GPT-5.6 Luna off 49.1%
Gemini 3.5 Flash-Lite off 49.1%
Muse Spark Open high 49.1%
GPT-5.6 Terra off 48.6%
DeepSeek V4.1 Flash off 48.6%
Kimi K3 off 46.3%
Gemini 3.5 Flash-Lite high 45%
Qwen3.8-2.4T-A95B on 45%
MiniMax-M3.1-Flash off 44.5%
Gemma 4 E2B off 44%
Kimi K3 on 44%
DeepSeek-V4-Pro on 43.6%
Gemini 2.5 Flash-Lite Sep 2025 off 42.7%
Gemini 2.5 Pro Jun 2025 high 41.3%
Inkling off 40.8%
Muse Spark Open off 40.8%
Inkling high 40.4%
MiniMax-M3 on 37.2%
Gemini 2.5 Pro Jun 2025 minimal 36.7%
Gemma 4 12B off 36.7%
Gemma 4 26B A4B off 36.2%
GPT-5.2 off 35.8%
GPT-5.2 high 35.3%
GLM-5.3-Flash on 35.3%
GPT-4.1 nano none 34.4%
GPT-4o Mar 2025 none 33.9%
Mistral Small 4 off 33.9%
Apertus 1.5 70B on 33.5%
Inkling-Small high 30.7%
Gemma 4 E2B on 29.8%
Gemma 4 E4B off 29.8%
Qwen 3.8 27B off 27.1%
Mistral Large 4 low 25.7%
Gemma 4 E4B on 25.2%
Mistral Large 4 high 25.2%
MiniMax-M3.1-Flash high 24.8%
Apertus 1.5 8B on 21.6%
Mistral Medium 3.5 off 20.2%
Gemini 2.5 Flash-Lite Sep 2025 high 19.7%
Gemma 4 31B IT on 19.7%
Qwen 3.8 27B xhigh 19.3%
Mistral Large 3 none 18.8%
GPT-3.5 Turbo Jan 2024 none 17.4%
Gemma 4 12B on 17.4%
Llama 3.1-70B none 17%
Gemma 4 26B A4B on 16.5%
Mistral Small 4 high 15.6%
Claude Haiku 4.5 off 14.7%
Claude Haiku 4.5 high 11.5%
Mistral Medium 3.5 high 10.6%
Ripple-65B Preview on 10.6%
Llama 3.1-8B none 8.26%
Apertus 8B Instruct none 2.29%

2 models have held number one on the official leaderboard since its release in September 2026

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

Anthropic
Own leaderboard 2 leaders

Source: Official Miðeind Icelandic Grammar Correction leaderboard

Scores over time

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% 2024 2025 2026 2027 Near the ceiling (85%) Claude Fable 5.2 Gemini 3.1 Pro Gemini 2.5 Flash-Lite GPT-4.1 nano GPT-3.5 Turbo 0% 20% 40% 60% 80% 100% 2024 2025 2026 2027 Near the ceiling (85%) Claude Fable 5.2 Gemini 3.1 Pro

Source: Official Miðeind Icelandic Grammar Correction 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 11 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access The tasks are private.
  • Pass Confidence Checked against a primary source.
Scores 5 pass
  • Pass Independent runs Scores from the official leaderboard (62 entries).
  • Pass Models scored 61 models we track, plus 1 other entry on its leaderboard.
  • Pass Newest score The newest is from November 2026, for Gemini 4 Neon.
  • Pass Frontier models 9 of the top 10 have a score, led by Claude Fable 5.2 (#2).
  • Pass Headroom The best is 83.9%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on mideind-icelandic-llm-leaderboard.static.hf.space: 103 rows for models we track and 1 other entry 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