Benchmark Almanac

Miðeind Icelandic Idioms

Claude Fable 5.2 85.7% leads since November 2026
Not in model cards

Models supply a missing word in an Icelandic idiom and explain its meaning. A program checks the missing word. A language model grades the explanation.

Passes test 2. Miðeind reports scores for 60 named models on its current leaderboard. The two tasks use the same idioms and count once in the average.

Made by
Miðeind
Released
September 2026
Kind
Reasoning in many languages
Level
Everyday
Scored
Graded by a model
Answers
Short answer
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 85.7% Claude Fable 5.2 leads with 85.7%

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 85.7%
  2. 2 Claude Opus 5.5 high 79.5%
  3. 3 Gemini 4 Argon high 78.8%
  4. 4 GPT-6 Astra low 78.5%
  5. 5 Claude Opus 5 high 77.6%
  6. 6 GPT-6.1 Sol high 77%
  7. 7 Claude Fable 5.1 high 76.8%
  8. 8 Gemini 4 Neon low 76.8%
  9. 9 Gemini 3.1 Pro high 76.1%
  10. 10 Claude Opus 4.8 high 76.1%
  11. 11 GPT-6.2 Sol high 75.7%
  12. 12 Gemini 3.7 Flash high 74.9%
  13. 13 Gemini 3.8 Flash high 74.9%
  14. 14 Gemini 4 Hydrogen low 74.1%
  15. 15 GPT-6 Sol high 73.7%
  16. 16 GPT-5.6 Sol high 73%
  17. 17 GPT-6.1 Terra high 73%
  18. 18 Gemini 4 Carbon low 71.6%
  19. 19 Gemini 3.5 Flash high 70.7%
  20. 20 GPT-6 Luna high 69.5%
  21. 21 Claude Sonnet 5.5 high 69.4%
  22. 22 GPT-5.6 Luna high 66%
  23. 23 Claude Sonnet 5 high 65%
  24. 24 Claude Haiku 5.5 high 63.5%
  25. 25 Gemini 3.5 Flash-Lite high 60.6%
  26. 26 GPT-5.6 Terra high 60.5%
  27. 27 Gemini 2.5 Pro Jun 2025, high 59.8%
  28. 28 Inkling 2 Small minimal 59.1%
  29. 29 Inkling high 57.4%
  30. 30 GPT-5.2 high 55.4%
  31. 31 DeepSeek-V4-Pro on 55.1%
  32. 32 Grok 4.7 low 53%
  33. 33 Muse Spark Open high 50.2%
  34. 34 DeepSeek V4.1 Flash on 48.4%
  35. 35 GPT-4o Mar 2025, none 46.4%
  36. 36 Qwen3.8-2.4T-A95B on 43.5%
  37. 37 Mistral Large 4 high 36.2%
  38. 38 Gemini 2.5 Flash-Lite Sep 2025, high 35.2%
  39. 39 MiniMax-M3 on 34.2%
  40. 40 Inkling-Small high 33.6%
  41. 41 Kimi K3 on 32.4%
  42. 42 Gemma 4 31B IT on 28.8%
  43. 43 GLM-5.3-Flash on 27.5%
  44. 44 Bielik 11B v3.1 Instruct high 25.8%
  45. 45 Gemma 4 26B A4B on 25.4%
  46. 46 Mistral Large 3 none 25%
  47. 47 Claude Haiku 4.5 high 23.5%
  48. 48 Mistral Small 4 high 19.1%
  49. 49 Qwen 3.8 27B xhigh 19%
  50. 50 Gemma 4 12B on 18.1%
  51. 51 Apertus 1.5 70B on 16.6%
  52. 52 Mistral Medium 3.5 high 15.8%
  53. 53 MiniMax-M3.1-Flash high 15.6%
  54. 54 GPT-4.1 nano none 14.6%
  55. 55 Llama 3.1-70B none 13.7%
  56. 56 Apertus 1.5 8B on 11.5%
  57. 57 Ripple-65B Preview on 9.80%
  58. 58 GPT-3.5 Turbo Jan 2024, none 8.91%
  59. 59 Apertus 8B Instruct none 7.67%
  60. 60 Gemma 4 E4B on 7.55%
  61. 61 Gemma 4 E2B on 4.90%
  62. 62 Llama 3.1-8B none 2.81%

Source: Official Miðeind Icelandic Idioms leaderboard

  1. 1 Claude Fable 5.2 high 85.7%
  2. 2 Claude Opus 5.5 high 79.5%
  3. 3 Gemini 4 Argon high 78.8%
  4. 4 GPT-6 Astra low 78.5%
  5. 5 Claude Opus 5 high 77.6%
  6. 6 GPT-6.1 Sol high 77%
  7. 7 Claude Fable 5.1 high 76.8%
  8. 8 Gemini 4 Neon low 76.8%
  9. 9 Gemini 3.1 Pro high 76.1%
  10. 10 Claude Opus 4.8 high 76.1%
  11. 11 GPT-6.2 Sol high 75.7%
  12. 12 Gemini 3.7 Flash high 74.9%
  13. 13 Gemini 3.8 Flash high 74.9%
  14. 14 Gemini 4 Hydrogen low 74.1%
  15. 15 GPT-6 Sol high 73.7%
  16. 16 GPT-5.6 Sol high 73%
  17. 17 GPT-6.1 Terra high 73%
  18. 18 Gemini 4 Carbon low 71.6%
  19. 19 Gemini 3.5 Flash high 70.7%
  20. 20 GPT-6 Luna high 69.5%
  21. 21 Claude Sonnet 5.5 high 69.4%
  22. 22 GPT-5.6 Luna high 66%
  23. 23 Claude Sonnet 5 high 65%
  24. 24 Claude Haiku 5.5 high 63.5%
  25. 25 Gemini 3.5 Flash-Lite high 60.6%
  26. 26 GPT-5.6 Terra high 60.5%
  27. 27 Gemini 2.5 Pro Jun 2025, high 59.8%
  28. 28 Inkling 2 Small minimal 59.1%
  29. 29 Inkling high 57.4%
  30. 30 GPT-5.2 high 55.4%
  31. 31 DeepSeek-V4-Pro on 55.1%
  32. 32 Grok 4.7 low 53%
  33. 33 Muse Spark Open high 50.2%
  34. 34 DeepSeek V4.1 Flash on 48.4%
  35. 35 GPT-4o Mar 2025, none 46.4%
  36. 36 Qwen3.8-2.4T-A95B on 43.5%
  37. 37 Mistral Large 4 high 36.2%
  38. 38 Gemini 2.5 Flash-Lite Sep 2025, high 35.2%
  39. 39 MiniMax-M3 on 34.2%
  40. 40 Inkling-Small high 33.6%
  41. 41 Kimi K3 on 32.4%
  42. 42 Gemma 4 31B IT on 28.8%
  43. 43 GLM-5.3-Flash on 27.5%
  44. 44 Gemma 4 26B A4B on 25.4%
  45. 45 Mistral Large 3 none 25%
  46. 46 Claude Haiku 4.5 high 23.5%
  47. 47 Mistral Small 4 high 19.1%
  48. 48 Qwen 3.8 27B xhigh 19%
  49. 49 Gemma 4 12B on 18.1%
  50. 50 Apertus 1.5 70B on 16.6%
  51. 51 Mistral Medium 3.5 high 15.8%
  52. 52 MiniMax-M3.1-Flash high 15.6%
  53. 53 GPT-4.1 nano none 14.6%
  54. 54 Llama 3.1-70B none 13.7%
  55. 55 Apertus 1.5 8B on 11.5%
  56. 56 Ripple-65B Preview on 9.80%
  57. 57 GPT-3.5 Turbo Jan 2024, none 8.91%
  58. 58 Apertus 8B Instruct none 7.67%
  59. 59 Gemma 4 E4B on 7.55%
  60. 60 Gemma 4 E2B on 4.90%
  61. 61 Llama 3.1-8B none 2.81%

Source: Official Miðeind Icelandic Idioms leaderboard

#EntryScoreSettingReleasedRun onLinkSource
1 Claude Fable 5.2 85.7% high Nov 2026 Nov 2026 anthropic.com Official leaderboard
2 Claude Opus 5.5 79.5% high Sep 2026 Sep 2026 anthropic.com Official leaderboard
3 Gemini 4 Argon 78.8% high Sep 2026 Oct 2026 ai.google.dev Official leaderboard
4 GPT-6 Astra 78.5% low Sep 2026 Sep 2026 developers.openai.com Official leaderboard
5 Claude Opus 5 77.6% high Jul 2026 anthropic.com Official leaderboard
6 GPT-6.1 Sol 77% high Sep 2026 Sep 2026 developers.openai.com Official leaderboard
7 Claude Fable 5.1 76.8% high Sep 2026 anthropic.com Official leaderboard
8 Gemini 4 Neon 76.8% low Nov 2026 Nov 2026 ai.google.dev Official leaderboard
9 Gemini 3.1 Pro 76.1% high Feb 2026 ai.google.dev Official leaderboard
10 Claude Opus 4.8 76.1% high May 2026 anthropic.com Official leaderboard
11 GPT-6.2 Sol 75.7% high Nov 2026 Nov 2026 developers.openai.com Official leaderboard
12 Gemini 3.7 Flash 74.9% high Aug 2026 ai.google.dev Official leaderboard
13 Gemini 3.8 Flash 74.9% high Sep 2026 Sep 2026 ai.google.dev Official leaderboard
14 Gemini 4 Hydrogen 74.1% low Nov 2026 Nov 2026 ai.google.dev Official leaderboard
15 GPT-6 Sol 73.7% high Sep 2026 Sep 2026 developers.openai.com Official leaderboard
16 GPT-5.6 Sol 73% high Jul 2026 developers.openai.com Official leaderboard
17 GPT-6.1 Terra 73% high Oct 2026 Oct 2026 developers.openai.com Official leaderboard
18 Gemini 4 Carbon 71.6% low Oct 2026 Nov 2026 ai.google.dev Official leaderboard
19 Gemini 3.5 Flash 70.7% high May 2026 ai.google.dev Official leaderboard
20 GPT-6 Luna 69.5% high Sep 2026 Sep 2026 developers.openai.com Official leaderboard
21 Claude Sonnet 5.5 69.4% high Sep 2026 Sep 2026 anthropic.com Official leaderboard
22 GPT-5.6 Luna 66% high Jul 2026 developers.openai.com Official leaderboard
23 Claude Sonnet 5 65% high Jun 2026 anthropic.com Official leaderboard
24 Claude Haiku 5.5 63.5% high Oct 2026 Oct 2026 anthropic.com Official leaderboard
25 Gemini 3.5 Flash-Lite 60.6% high Jul 2026 ai.google.dev Official leaderboard
26 GPT-5.6 Terra 60.5% high Jul 2026 developers.openai.com Official leaderboard
27 Gemini 2.5 Pro Jun 2025 59.8% high Jun 2025 Sep 2026 ai.google.dev Official leaderboard
28 Inkling 2 Small 59.1% minimal Nov 2026 Nov 2026 huggingface.co Official leaderboard
29 Inkling 57.4% high Jul 2026 Sep 2026 huggingface.co Official leaderboard
30 GPT-5.2 55.4% high Dec 2025 platform.openai.com Official leaderboard
31 DeepSeek-V4-Pro 55.1% on Apr 2026 huggingface.co Official leaderboard
32 Grok 4.7 53% low Sep 2026 Nov 2026 docs.x.ai Official leaderboard
33 Muse Spark Open 50.2% high Oct 2026 Oct 2026 huggingface.co Official leaderboard
34 DeepSeek V4.1 Flash 48.4% on Sep 2026 Oct 2026 huggingface.co Official leaderboard
35 GPT-4o Mar 2025 46.4% none Mar 2025 Sep 2026 platform.openai.com Official leaderboard
36 Qwen3.8-2.4T-A95B 43.5% on Aug 2026 Sep 2026 huggingface.co Official leaderboard
37 Mistral Large 4 36.2% high Oct 2026 Oct 2026 mistral.ai Official leaderboard
38 Gemini 2.5 Flash-Lite Sep 2025 35.2% high Sep 2025 ai.google.dev Official leaderboard
39 MiniMax-M3 34.2% on Jun 2026 huggingface.co Official leaderboard
40 Inkling-Small 33.6% high Jul 2026 Sep 2026 huggingface.co Official leaderboard
41 Kimi K3 32.4% on Jul 2026 Sep 2026 huggingface.co Official leaderboard
42 Gemma 4 31B IT 28.8% on Apr 2026 Sep 2026 deepmind.google Official leaderboard
43 GLM-5.3-Flash 27.5% on Aug 2026 huggingface.co Official leaderboard
44 Bielik 11B v3.1 Instruct 11.2B 25.8% high Nov 2026 huggingface.co Official leaderboard
45 Gemma 4 26B A4B 25.4% on Apr 2026 huggingface.co Official leaderboard
46 Mistral Large 3 25% none Dec 2025 Sep 2026 mistral.ai Official leaderboard
47 Claude Haiku 4.5 23.5% high Oct 2025 anthropic.com Official leaderboard
48 Mistral Small 4 19.1% high Mar 2026 Sep 2026 mistral.ai Official leaderboard
49 Qwen 3.8 27B 19% xhigh Aug 2026 Oct 2026 huggingface.co Official leaderboard
50 Gemma 4 12B 18.1% on Jun 2026 Nov 2026 huggingface.co Official leaderboard
51 Apertus 1.5 70B 16.6% on Jul 2026 Sep 2026 huggingface.co Official leaderboard
52 Mistral Medium 3.5 15.8% high Apr 2026 Sep 2026 mistral.ai Official leaderboard
53 MiniMax-M3.1-Flash 15.6% high Oct 2026 Oct 2026 huggingface.co Official leaderboard
54 GPT-4.1 nano 14.6% none Apr 2025 platform.openai.com Official leaderboard
55 Llama 3.1-70B 13.7% none Jul 2024 Oct 2026 huggingface.co Official leaderboard
56 Apertus 1.5 8B 11.5% on Jul 2026 Sep 2026 huggingface.co Official leaderboard
57 Ripple-65B Preview 9.80% on Nov 2026 Nov 2026 huggingface.co Official leaderboard
58 GPT-3.5 Turbo Jan 2024 8.91% none Jan 2024 Sep 2026 platform.openai.com Official leaderboard
59 Apertus 8B Instruct 7.67% none Sep 2025 Sep 2026 huggingface.co Official leaderboard
60 Gemma 4 E4B 7.55% on Mar 2026 Oct 2026 huggingface.co Official leaderboard
61 Gemma 4 E2B 4.90% on Mar 2026 Oct 2026 huggingface.co Official leaderboard
62 Llama 3.1-8B 2.81% none Jul 2024 Oct 2026 huggingface.co Official leaderboard
#ModelScoreSettingReleasedRun onLinkSource
1 Claude Fable 5.2 85.7% high Nov 2026 Nov 2026 anthropic.com Official leaderboard
2 Claude Opus 5.5 79.5% high Sep 2026 Sep 2026 anthropic.com Official leaderboard
3 Gemini 4 Argon 78.8% high Sep 2026 Oct 2026 ai.google.dev Official leaderboard
4 GPT-6 Astra 78.5% low Sep 2026 Sep 2026 developers.openai.com Official leaderboard
5 Claude Opus 5 77.6% high Jul 2026 anthropic.com Official leaderboard
6 GPT-6.1 Sol 77% high Sep 2026 Sep 2026 developers.openai.com Official leaderboard
7 Claude Fable 5.1 76.8% high Sep 2026 anthropic.com Official leaderboard
8 Gemini 4 Neon 76.8% low Nov 2026 Nov 2026 ai.google.dev Official leaderboard
9 Gemini 3.1 Pro 76.1% high Feb 2026 ai.google.dev Official leaderboard
10 Claude Opus 4.8 76.1% high May 2026 anthropic.com Official leaderboard
11 GPT-6.2 Sol 75.7% high Nov 2026 Nov 2026 developers.openai.com Official leaderboard
12 Gemini 3.7 Flash 74.9% high Aug 2026 ai.google.dev Official leaderboard
13 Gemini 3.8 Flash 74.9% high Sep 2026 Sep 2026 ai.google.dev Official leaderboard
14 Gemini 4 Hydrogen 74.1% low Nov 2026 Nov 2026 ai.google.dev Official leaderboard
15 GPT-6 Sol 73.7% high Sep 2026 Sep 2026 developers.openai.com Official leaderboard
16 GPT-5.6 Sol 73% high Jul 2026 developers.openai.com Official leaderboard
17 GPT-6.1 Terra 73% high Oct 2026 Oct 2026 developers.openai.com Official leaderboard
18 Gemini 4 Carbon 71.6% low Oct 2026 Nov 2026 ai.google.dev Official leaderboard
19 Gemini 3.5 Flash 70.7% high May 2026 ai.google.dev Official leaderboard
20 GPT-6 Luna 69.5% high Sep 2026 Sep 2026 developers.openai.com Official leaderboard
21 Claude Sonnet 5.5 69.4% high Sep 2026 Sep 2026 anthropic.com Official leaderboard
22 GPT-5.6 Luna 66% high Jul 2026 developers.openai.com Official leaderboard
23 Claude Sonnet 5 65% high Jun 2026 anthropic.com Official leaderboard
24 Claude Haiku 5.5 63.5% high Oct 2026 Oct 2026 anthropic.com Official leaderboard
25 Gemini 3.5 Flash-Lite 60.6% high Jul 2026 ai.google.dev Official leaderboard
26 GPT-5.6 Terra 60.5% high Jul 2026 developers.openai.com Official leaderboard
27 Gemini 2.5 Pro Jun 2025 59.8% high Jun 2025 Sep 2026 ai.google.dev Official leaderboard
28 Inkling 2 Small 59.1% minimal Nov 2026 Nov 2026 huggingface.co Official leaderboard
29 Inkling 57.4% high Jul 2026 Sep 2026 huggingface.co Official leaderboard
30 GPT-5.2 55.4% high Dec 2025 platform.openai.com Official leaderboard
31 DeepSeek-V4-Pro 55.1% on Apr 2026 huggingface.co Official leaderboard
32 Grok 4.7 53% low Sep 2026 Nov 2026 docs.x.ai Official leaderboard
33 Muse Spark Open 50.2% high Oct 2026 Oct 2026 huggingface.co Official leaderboard
34 DeepSeek V4.1 Flash 48.4% on Sep 2026 Oct 2026 huggingface.co Official leaderboard
35 GPT-4o Mar 2025 46.4% none Mar 2025 Sep 2026 platform.openai.com Official leaderboard
36 Qwen3.8-2.4T-A95B 43.5% on Aug 2026 Sep 2026 huggingface.co Official leaderboard
37 Mistral Large 4 36.2% high Oct 2026 Oct 2026 mistral.ai Official leaderboard
38 Gemini 2.5 Flash-Lite Sep 2025 35.2% high Sep 2025 ai.google.dev Official leaderboard
39 MiniMax-M3 34.2% on Jun 2026 huggingface.co Official leaderboard
40 Inkling-Small 33.6% high Jul 2026 Sep 2026 huggingface.co Official leaderboard
41 Kimi K3 32.4% on Jul 2026 Sep 2026 huggingface.co Official leaderboard
42 Gemma 4 31B IT 28.8% on Apr 2026 Sep 2026 deepmind.google Official leaderboard
43 GLM-5.3-Flash 27.5% on Aug 2026 huggingface.co Official leaderboard
44 Gemma 4 26B A4B 25.4% on Apr 2026 huggingface.co Official leaderboard
45 Mistral Large 3 25% none Dec 2025 Sep 2026 mistral.ai Official leaderboard
46 Claude Haiku 4.5 23.5% high Oct 2025 anthropic.com Official leaderboard
47 Mistral Small 4 19.1% high Mar 2026 Sep 2026 mistral.ai Official leaderboard
48 Qwen 3.8 27B 19% xhigh Aug 2026 Oct 2026 huggingface.co Official leaderboard
49 Gemma 4 12B 18.1% on Jun 2026 Nov 2026 huggingface.co Official leaderboard
50 Apertus 1.5 70B 16.6% on Jul 2026 Sep 2026 huggingface.co Official leaderboard
51 Mistral Medium 3.5 15.8% high Apr 2026 Sep 2026 mistral.ai Official leaderboard
52 MiniMax-M3.1-Flash 15.6% high Oct 2026 Oct 2026 huggingface.co Official leaderboard
53 GPT-4.1 nano 14.6% none Apr 2025 platform.openai.com Official leaderboard
54 Llama 3.1-70B 13.7% none Jul 2024 Oct 2026 huggingface.co Official leaderboard
55 Apertus 1.5 8B 11.5% on Jul 2026 Sep 2026 huggingface.co Official leaderboard
56 Ripple-65B Preview 9.80% on Nov 2026 Nov 2026 huggingface.co Official leaderboard
57 GPT-3.5 Turbo Jan 2024 8.91% none Jan 2024 Sep 2026 platform.openai.com Official leaderboard
58 Apertus 8B Instruct 7.67% none Sep 2025 Sep 2026 huggingface.co Official leaderboard
59 Gemma 4 E4B 7.55% on Mar 2026 Oct 2026 huggingface.co Official leaderboard
60 Gemma 4 E2B 4.90% on Mar 2026 Oct 2026 huggingface.co Official leaderboard
61 Llama 3.1-8B 2.81% none Jul 2024 Oct 2026 huggingface.co Official leaderboard

Rankings by part

We read 2 parts from the benchmark’s own leaderboard: Completion and Meanings. Each ranks models on one part of the test. Claude Fable 5.2 leads both. Arrows show how far a model moved from its overall rank among the models we track.

Completion

61 models we track have a score here.

  1. 1 Claude Fable 5.2 83.4%
  2. 2 Gemini 4 Argon ↑1 76%
  3. 3 Claude Opus 5.5 ↓1 75%
  4. 4 GPT-6 Astra 72.5%
  5. 5 GPT-6.1 Sol ↑1 71.2%
  6. 6 Gemini 4 Neon ↑2 71.2%
  7. 7 Claude Opus 5 ↓2 70.7%
  8. 8 Gemini 3.1 Pro ↑1 70.4%
  9. 9 Claude Fable 5.1 ↓2 70.1%
  10. 10 Gemini 4 Hydrogen ↑4 69.6%

More effort adds a median 1 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
85.7%
Gemini 4 Argon
off high
78.8%
GPT-6 Astra
low
78.5%
Claude Opus 5
off high
77.6%
GPT-6.1 Sol
low
77%
Gemini 3.1 Pro
low
76.1%
Claude Opus 4.8
off high
76.1%
GPT-6.2 Sol
low high
75.7%
Gemini 3.7 Flash
low
74.9%
Gemini 3.8 Flash
low high
74.9%
Gemini 4 Hydrogen
low
74.1%
GPT-6 Sol
off high
73.7%

Source: Official Miðeind Icelandic Idioms leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude Fable 5.2 high 85.7%
Claude Fable 5.2 low 84.9%
Claude Opus 5.5 high 79.5%
Gemini 4 Argon high 78.8%
GPT-6 Astra low 78.5%
GPT-6 Astra high 77.8%
Claude Opus 5 high 77.6%
GPT-6.1 Sol high 77%
Claude Fable 5.1 high 76.8%
Gemini 4 Neon low 76.8%
GPT-6.1 Sol low 76.4%
Gemini 3.1 Pro high 76.1%
Claude Opus 4.8 high 76.1%
GPT-6.2 Sol high 75.7%
Gemini 3.7 Flash high 74.9%
Gemini 3.8 Flash high 74.9%
Gemini 4 Hydrogen low 74.1%
Gemini 4 Hydrogen high 73.9%
Gemini 3.1 Pro low 73.8%
GPT-6 Sol high 73.7%
Gemini 3.7 Flash low 73.4%
GPT-5.6 Sol high 73%
GPT-6.1 Terra high 73%
Gemini 3.8 Flash low 72%
Gemini 4 Carbon low 71.6%
Gemini 3.5 Flash high 70.7%
Gemini 4 Argon off 70.4%
GPT-6.2 Sol low 69.7%
GPT-6 Luna high 69.5%
Claude Sonnet 5.5 high 69.4%
GPT-5.6 Luna high 66%
Claude Sonnet 5 high 65%
Claude Sonnet 5 off 64.1%
Claude Sonnet 5.5 off 64.1%
GPT-6.1 Terra off 63.8%
Claude Haiku 5.5 high 63.5%
GPT-5.6 Sol off 63%
GPT-6 Sol off 63%
Gemini 3.5 Flash off 61.7%
Claude Opus 5 off 61.6%
Gemini 3.5 Flash-Lite high 60.6%
GPT-5.6 Terra high 60.5%
Gemini 2.5 Pro Jun 2025 high 59.8%
Inkling 2 Small minimal 59.1%
Claude Opus 4.8 off 57.5%
Inkling 2 Small high 57.4%
Inkling high 57.4%
GPT-5.2 high 55.4%
DeepSeek-V4-Pro on 55.1%
Grok 4.7 low 53%
Gemini 2.5 Pro Jun 2025 minimal 52.9%
Claude Haiku 5.5 off 52.8%
Grok 4.7 high 52.6%
GPT-6 Luna off 50.3%
Muse Spark Open high 50.2%
GPT-5.6 Terra off 48.6%
DeepSeek V4.1 Flash on 48.4%
Inkling off 46.7%
GPT-4o Mar 2025 none 46.4%
GPT-5.6 Luna off 46.3%
GPT-5.2 off 46.1%
Gemini 3.5 Flash-Lite off 45.1%
Qwen3.8-2.4T-A95B on 43.5%
DeepSeek-V4-Pro off 42.1%
Mistral Large 4 high 36.2%
Gemini 2.5 Flash-Lite Sep 2025 high 35.2%
Muse Spark Open off 35%
Mistral Large 4 low 34.9%
DeepSeek V4.1 Flash off 34.9%
MiniMax-M3 on 34.2%
Inkling-Small high 33.6%
Kimi K3 on 32.4%
Gemini 2.5 Flash-Lite Sep 2025 off 30.1%
Gemma 4 31B IT on 28.8%
GLM-5.3-Flash on 27.5%
Gemma 4 26B A4B on 25.4%
Mistral Large 3 none 25%
Claude Haiku 4.5 high 23.5%
Kimi K3 off 23.5%
Claude Haiku 4.5 off 20.7%
Gemma 4 26B A4B off 19.7%
Mistral Small 4 high 19.1%
Qwen 3.8 27B xhigh 19%
Gemma 4 12B on 18.1%
Mistral Small 4 off 18.1%
Apertus 1.5 70B on 16.6%
Mistral Medium 3.5 high 15.8%
MiniMax-M3.1-Flash high 15.6%
Mistral Medium 3.5 off 15%
GPT-4.1 nano none 14.6%
Llama 3.1-70B none 13.7%
Apertus 1.5 8B on 11.5%
Qwen 3.8 27B off 11.3%
Gemma 4 12B off 10%
Ripple-65B Preview on 9.80%
GPT-3.5 Turbo Jan 2024 none 8.91%
Apertus 8B Instruct none 7.67%
Gemma 4 E4B on 7.55%
Gemma 4 E4B off 6.43%
MiniMax-M3.1-Flash off 5.12%
Gemma 4 E2B on 4.90%
Llama 3.1-8B none 2.81%
Gemma 4 E2B off 2.78%

More effort adds a median 1 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
85.7%
Gemini 4 Argon
off high
78.8%
GPT-6 Astra
low
78.5%
Claude Opus 5
off high
77.6%
GPT-6.1 Sol
low
77%
Gemini 3.1 Pro
low
76.1%
Claude Opus 4.8
off high
76.1%
GPT-6.2 Sol
low high
75.7%
Gemini 3.7 Flash
low
74.9%
Gemini 3.8 Flash
low high
74.9%
Gemini 4 Hydrogen
low
74.1%
GPT-6 Sol
off high
73.7%

Source: Official Miðeind Icelandic Idioms leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude Fable 5.2 high 85.7%
Claude Fable 5.2 low 84.9%
Claude Opus 5.5 high 79.5%
Gemini 4 Argon high 78.8%
GPT-6 Astra low 78.5%
GPT-6 Astra high 77.8%
Claude Opus 5 high 77.6%
GPT-6.1 Sol high 77%
Claude Fable 5.1 high 76.8%
Gemini 4 Neon low 76.8%
GPT-6.1 Sol low 76.4%
Gemini 3.1 Pro high 76.1%
Claude Opus 4.8 high 76.1%
GPT-6.2 Sol high 75.7%
Gemini 3.7 Flash high 74.9%
Gemini 3.8 Flash high 74.9%
Gemini 4 Hydrogen low 74.1%
Gemini 4 Hydrogen high 73.9%
Gemini 3.1 Pro low 73.8%
GPT-6 Sol high 73.7%
Gemini 3.7 Flash low 73.4%
GPT-5.6 Sol high 73%
GPT-6.1 Terra high 73%
Gemini 3.8 Flash low 72%
Gemini 4 Carbon low 71.6%
Gemini 3.5 Flash high 70.7%
Gemini 4 Argon off 70.4%
GPT-6.2 Sol low 69.7%
GPT-6 Luna high 69.5%
Claude Sonnet 5.5 high 69.4%
GPT-5.6 Luna high 66%
Claude Sonnet 5 high 65%
Claude Sonnet 5 off 64.1%
Claude Sonnet 5.5 off 64.1%
GPT-6.1 Terra off 63.8%
Claude Haiku 5.5 high 63.5%
GPT-5.6 Sol off 63%
GPT-6 Sol off 63%
Gemini 3.5 Flash off 61.7%
Claude Opus 5 off 61.6%
Gemini 3.5 Flash-Lite high 60.6%
GPT-5.6 Terra high 60.5%
Gemini 2.5 Pro Jun 2025 high 59.8%
Inkling 2 Small minimal 59.1%
Claude Opus 4.8 off 57.5%
Inkling 2 Small high 57.4%
Inkling high 57.4%
GPT-5.2 high 55.4%
DeepSeek-V4-Pro on 55.1%
Grok 4.7 low 53%
Gemini 2.5 Pro Jun 2025 minimal 52.9%
Claude Haiku 5.5 off 52.8%
Grok 4.7 high 52.6%
GPT-6 Luna off 50.3%
Muse Spark Open high 50.2%
GPT-5.6 Terra off 48.6%
DeepSeek V4.1 Flash on 48.4%
Inkling off 46.7%
GPT-4o Mar 2025 none 46.4%
GPT-5.6 Luna off 46.3%
GPT-5.2 off 46.1%
Gemini 3.5 Flash-Lite off 45.1%
Qwen3.8-2.4T-A95B on 43.5%
DeepSeek-V4-Pro off 42.1%
Mistral Large 4 high 36.2%
Gemini 2.5 Flash-Lite Sep 2025 high 35.2%
Muse Spark Open off 35%
Mistral Large 4 low 34.9%
DeepSeek V4.1 Flash off 34.9%
MiniMax-M3 on 34.2%
Inkling-Small high 33.6%
Kimi K3 on 32.4%
Gemini 2.5 Flash-Lite Sep 2025 off 30.1%
Gemma 4 31B IT on 28.8%
GLM-5.3-Flash on 27.5%
Gemma 4 26B A4B on 25.4%
Mistral Large 3 none 25%
Claude Haiku 4.5 high 23.5%
Kimi K3 off 23.5%
Claude Haiku 4.5 off 20.7%
Gemma 4 26B A4B off 19.7%
Mistral Small 4 high 19.1%
Qwen 3.8 27B xhigh 19%
Gemma 4 12B on 18.1%
Mistral Small 4 off 18.1%
Apertus 1.5 70B on 16.6%
Mistral Medium 3.5 high 15.8%
MiniMax-M3.1-Flash high 15.6%
Mistral Medium 3.5 off 15%
GPT-4.1 nano none 14.6%
Llama 3.1-70B none 13.7%
Apertus 1.5 8B on 11.5%
Qwen 3.8 27B off 11.3%
Gemma 4 12B off 10%
Ripple-65B Preview on 9.80%
GPT-3.5 Turbo Jan 2024 none 8.91%
Apertus 8B Instruct none 7.67%
Gemma 4 E4B on 7.55%
Gemma 4 E4B off 6.43%
MiniMax-M3.1-Flash off 5.12%
Gemma 4 E2B on 4.90%
Llama 3.1-8B none 2.81%
Gemma 4 E2B off 2.78%

4 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 OpenAI
Own leaderboard Passed 85% in Nov 2026, by Claude Fable 5.2

Source: Official Miðeind Icelandic Idioms leaderboard

Claude Fable 5.2 on the official leaderboard passed 85% in November 2026

That was 2 months after its release. By the same age, 21% of the benchmarks from 2026 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% 2024 2025 2026 2027 Near the ceiling (85%) Claude Fable 5.2 Claude Opus 5.5 Gemini 3.1 Pro Gemini 2.5 Pro GPT-4o Llama 3.1-70B GPT-3.5 Turbo 0% 20% 40% 60% 80% 100% 2024 2025 2026 2027 Near the ceiling (85%) Claude Fable 5.2 GPT-6 Astra

Source: Official Miðeind Icelandic Idioms 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 10 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access The tasks are private.
  • Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
  • 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).
  • Note Headroom Saturated. The best is 85.7%, past 85% since November 2026.
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