Benchmark Almanac

GreekMMLU

Gemini 3 Flash 93.2% leads since February 2026
Not in model cards

Models answer questions originally written in Greek for academic, professional, and government exams across 45 subjects.

EuroEval reports the Matthews correlation coefficient for its multiple-choice version.

Made by
Yang Zhang and colleagues
Released
February 2026
Kind
Knowledge in many languages
Level
University
Scored
Checked automatically
Answers
Multiple choice
Tasks
All public
Measures
Capability
Language
Greek
In and out
Text → Text
Interaction
Single turn
Items from
Exams and contests
Refresh
One fixed set
Run by
Anyone
In views
Law and finance
Strongest model tested
GPT-5.2 #51 on the capability index
Last model added
December 2025
Scores refreshed
November 20th, 2026
73
Entries in its paper, 34 models we track

Gemini 3 Flash leads with 93.2% among models released by December 2025 Gemini 3 Flash leads with 93.2% among models released by December 2025

Best score per entry from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. 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.

The newest model with a score here came out in December 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

73 entries. Scroll sideways for the rest.

  1. 93.2 Gemini 3 Flash
  2. 89.2 GPT-5.2
  3. 87.8 GPT-4o (snapshot unspecified)
  4. 82.6 Gemma 3 27B PT
  5. 82.2 Qwen2.5-72B
  6. 81.5 Llama 3.3 70B Instruct
  7. 81.3 Gemma 3 27B
  8. 80.8 Qwen3-30B-Instruct
  9. 80.7 Llama-3.1-70B-Instruct
  10. 80.4 Llama 3.1-70B
  11. 79.9 Qwen3-30B
  12. 77.3 Gemma 3 12B PT
  13. 77.3 Qwen3-14B
  14. 76.3 Gemma 3 12B
  15. 76.0 Qwen2.5-32B-Instruct
  16. 75.7 Qwen2.5-32B
  17. 75.0 EuroLLM-22B-Instruct-2512
  18. 74.1 EuroLLM-22B
  19. 72.8 Qwen3-8B
  20. 71.1 Qwen2.5-14B
  21. 70.9 Llama-Krikri-8B-Base
  22. 70.8 Llama-Krikri-8B-Instruct
  23. 70.8 Plutus-8B-instruct
  24. 70.6 Qwen2.5-14B-Instruct
  25. 70.5 Mistral Small 3
  26. 70.2 GLM-4-9B
  27. 68.8 GLM-4-9B-Chat
  28. 68.5 EuroLLM-9B-Instruct
  29. 68.3 qwen3-4b-instruct-2507
  30. 67.1 Qwen3-4B
  31. 66.3 EuroLLM-9B
  32. 65.6 Aya Expanse 8B
  33. 64.9 Llama-3-8B-Instruct
  34. 64.8 Gemma 3 4B
  35. 64.7 Llama-3.1-8B-Instruct
  36. 64.5 Gemma 3 4B PT
  37. 64.2 Llama 3-8B
  38. 64.2 Qwen2.5-7B-Instruct
  39. 63.7 Meltemi-7B-Instruct-v1.5
  40. 63.3 Llama 3.1-8B
  41. 63.0 Meltemi-7B-v1.5
  42. 63.0 Qwen2.5-7B
  43. 58.6 Aya-101
  44. 58.4 Meltemi-7B-v1
  45. 58.0 Qwen3-1.7B
  46. 56.6 mT0-xxl
  47. 53.4 Qwen2.5-3B-Instruct
  48. 51.6 Mistral-7B-Instruct-v0.3
  49. 50.5 Llama-3.2-3B-Instruct
  50. 50.0 mT0-xl
  51. 48.4 Llama 3.2 3B
  52. 47.5 Qwen2.5-3B
  53. 46.5 Qwen2.5-1.5B-Instruct
  54. 45.7 Gemma 3 1B
  55. 42.9 Qwen2.5-1.5B
  56. 41.0 Qwen3-0.6B
  57. 40.5 Llama-3.2-1B-Instruct
  58. 37.0 mT0-large
  59. 36.8 Llama-2-7b-chat-hf
  60. 36.4 Llama 3.2 1B
  61. 35.3 Llama 2-7B
  62. 35.1 BLOOMZ-1b1
  63. 34.9 Qwen2.5-0.5B-Instruct
  64. 34.7 Qwen2.5-0.5B
  65. 34.7 BLOOMZ-1b7
  66. 33.0 XGLM-4.5B
  67. 33.0 XGLM-1.7B
  68. 32.7 XGLM-2.9B
  69. 32.4 BLOOMZ-7b1
  70. 32.4 EuroLLM-1.7B-Instruct
  71. 32.3 EuroLLM-1.7B
  72. 32.2 XGLM-7.5B
  73. 31.9 Gemma 3 1B PT
  1. 1 Gemini 3 Flash zero-shot 93.2%
  2. 2 GPT-5.2 five-shot 89.2%
  3. 3 GPT-4o (snapshot unspecified) five-shot 87.8%
  4. 4 Gemma 3 27B PT five-shot 82.6%
  5. 5 Qwen2.5-72B five-shot 82.2%
  6. 6 Llama 3.3 70B Instruct five-shot 81.5%
  7. 7 Gemma 3 27B five-shot 81.3%
  8. 8 Qwen3-30B-Instruct five-shot 80.8%
  9. 9 Llama-3.1-70B-Instruct five-shot 80.7%
  10. 10 Llama 3.1-70B five-shot 80.4%
  11. 11 Qwen3-30B five-shot 79.9%
  12. 12 Gemma 3 12B PT five-shot 77.3%
  13. 13 Qwen3-14B five-shot 77.3%
  14. 14 Gemma 3 12B five-shot 76.3%
  15. 15 Qwen2.5-32B-Instruct five-shot 76%
  16. 16 Qwen2.5-32B five-shot 75.7%
  17. 17 EuroLLM-22B-Instruct-2512 five-shot 75%
  18. 18 EuroLLM-22B five-shot 74.1%
  19. 19 Qwen3-8B five-shot 72.8%
  20. 20 Qwen2.5-14B five-shot 71.1%
  21. 21 Llama-Krikri-8B-Base five-shot 70.9%
  22. 22 Llama-Krikri-8B-Instruct five-shot 70.8%
  23. 23 Plutus-8B-instruct five-shot 70.8%
  24. 24 Qwen2.5-14B-Instruct five-shot 70.6%
  25. 25 Mistral Small 3 zero-shot 70.5%
  26. 26 GLM-4-9B five-shot 70.2%
  27. 27 GLM-4-9B-Chat five-shot 68.8%
  28. 28 EuroLLM-9B-Instruct zero-shot 68.5%
  29. 29 qwen3-4b-instruct-2507 five-shot 68.3%
  30. 30 Qwen3-4B five-shot 67.1%
  31. 31 EuroLLM-9B zero-shot 66.3%
  32. 32 Aya Expanse 8B five-shot 65.6%
  33. 33 Llama-3-8B-Instruct five-shot 64.9%
  34. 34 Gemma 3 4B five-shot 64.8%
  35. 35 Llama-3.1-8B-Instruct five-shot 64.7%
  36. 36 Gemma 3 4B PT five-shot 64.5%
  37. 37 Llama 3-8B five-shot 64.2%
  38. 38 Qwen2.5-7B-Instruct five-shot 64.2%
  39. 39 Meltemi-7B-Instruct-v1.5 five-shot 63.7%
  40. 40 Llama 3.1-8B five-shot 63.3%
  41. 41 Meltemi-7B-v1.5 five-shot 63%
  42. 42 Qwen2.5-7B five-shot 63%
  43. 43 Aya-101 five-shot 58.6%
  44. 44 Meltemi-7B-v1 five-shot 58.4%
  45. 45 Qwen3-1.7B five-shot 58%
  46. 46 mT0-xxl zero-shot 56.6%
  47. 47 Qwen2.5-3B-Instruct five-shot 53.4%
  48. 48 Mistral-7B-Instruct-v0.3 five-shot 51.6%
  49. 49 Llama-3.2-3B-Instruct five-shot 50.5%
  50. 50 mT0-xl zero-shot 50%
  51. 51 Llama 3.2 3B five-shot 48.4%
  52. 52 Qwen2.5-3B five-shot 47.5%
  53. 53 Qwen2.5-1.5B-Instruct five-shot 46.5%
  54. 54 Gemma 3 1B five-shot 45.7%
  55. 55 Qwen2.5-1.5B five-shot 42.9%
  56. 56 Qwen3-0.6B five-shot 41%
  57. 57 Llama-3.2-1B-Instruct five-shot 40.5%
  58. 58 mT0-large zero-shot 37%
  59. 59 Llama-2-7b-chat-hf five-shot 36.8%
  60. 60 Llama 3.2 1B zero-shot 36.4%
  61. 61 Llama 2-7B zero-shot 35.3%
  62. 62 BLOOMZ-1b1 zero-shot 35.1%
  63. 63 Qwen2.5-0.5B-Instruct zero-shot 34.9%
  64. 64 Qwen2.5-0.5B zero-shot 34.7%
  65. 65 BLOOMZ-1b7 zero-shot 34.7%
  66. 66 XGLM-4.5B zero-shot 33%
  67. 67 XGLM-1.7B zero-shot 33%
  68. 68 XGLM-2.9B zero-shot 32.7%
  69. 69 BLOOMZ-7b1 zero-shot 32.4%
  70. 70 EuroLLM-1.7B-Instruct five-shot 32.4%
  71. 71 EuroLLM-1.7B zero-shot 32.3%
  72. 72 XGLM-7.5B zero-shot 32.2%
  73. 73 Gemma 3 1B PT zero-shot 31.9%

Source: The GreekMMLU paper

Source: The GreekMMLU paper

#EntryScoreSettingDateSource
1 Gemini 3 Flash 93.2% zero-shot Dec 2025 Official leaderboard
2 GPT-5.2 89.2% five-shot Dec 2025 Official leaderboard
3 GPT-4o (snapshot unspecified) 87.8% five-shot Official leaderboard
4 Gemma 3 27B PT 82.6% five-shot Mar 2025 Official leaderboard
5 Qwen2.5-72B 82.2% five-shot Sep 2024 Official leaderboard
6 Llama 3.3 70B Instruct 81.5% five-shot Dec 2024 Official leaderboard
7 Gemma 3 27B 81.3% five-shot Mar 2025 Official leaderboard
8 Qwen3-30B-Instruct 80.8% five-shot Official leaderboard
9 Llama-3.1-70B-Instruct 80.7% five-shot Official leaderboard
10 Llama 3.1-70B 80.4% five-shot Jul 2024 Official leaderboard
11 Qwen3-30B 79.9% five-shot Official leaderboard
12 Gemma 3 12B PT 77.3% five-shot Mar 2025 Official leaderboard
13 Qwen3-14B 77.3% five-shot Apr 2025 Official leaderboard
14 Gemma 3 12B 76.3% five-shot Mar 2025 Official leaderboard
15 Qwen2.5-32B-Instruct 76% five-shot Official leaderboard
16 Qwen2.5-32B 75.7% five-shot Sep 2024 Official leaderboard
17 EuroLLM-22B-Instruct-2512 75% five-shot Official leaderboard
18 EuroLLM-22B 74.1% five-shot Official leaderboard
19 Qwen3-8B 72.8% five-shot Apr 2025 Official leaderboard
20 Qwen2.5-14B 71.1% five-shot Sep 2024 Official leaderboard
21 Llama-Krikri-8B-Base 70.9% five-shot Official leaderboard
22 Llama-Krikri-8B-Instruct 70.8% five-shot Official leaderboard
23 Plutus-8B-instruct 70.8% five-shot Official leaderboard
24 Qwen2.5-14B-Instruct 70.6% five-shot Official leaderboard
25 Mistral Small 3 70.5% zero-shot Jan 2025 Official leaderboard
26 GLM-4-9B 70.2% five-shot Jun 2024 Official leaderboard
27 GLM-4-9B-Chat 68.8% five-shot Jun 2024 Official leaderboard
28 EuroLLM-9B-Instruct 68.5% zero-shot Official leaderboard
29 qwen3-4b-instruct-2507 68.3% five-shot Aug 2025 Official leaderboard
30 Qwen3-4B 67.1% five-shot Apr 2025 Official leaderboard
31 EuroLLM-9B 66.3% zero-shot Official leaderboard
32 Aya Expanse 8B 65.6% five-shot Oct 2024 Official leaderboard
33 Llama-3-8B-Instruct 64.9% five-shot Official leaderboard
34 Gemma 3 4B 64.8% five-shot Mar 2025 Official leaderboard
35 Llama-3.1-8B-Instruct 64.7% five-shot Official leaderboard
36 Gemma 3 4B PT 64.5% five-shot Mar 2025 Official leaderboard
37 Llama 3-8B 64.2% five-shot Apr 2024 Official leaderboard
38 Qwen2.5-7B-Instruct 64.2% five-shot Official leaderboard
39 Meltemi-7B-Instruct-v1.5 63.7% five-shot Official leaderboard
40 Llama 3.1-8B 63.3% five-shot Jul 2024 Official leaderboard
41 Meltemi-7B-v1.5 63% five-shot Official leaderboard
42 Qwen2.5-7B 63% five-shot Sep 2024 Official leaderboard
43 Aya-101 58.6% five-shot Official leaderboard
44 Meltemi-7B-v1 58.4% five-shot Official leaderboard
45 Qwen3-1.7B 58% five-shot Apr 2025 Official leaderboard
46 mT0-xxl 56.6% zero-shot Official leaderboard
47 Qwen2.5-3B-Instruct 53.4% five-shot Official leaderboard
48 Mistral-7B-Instruct-v0.3 51.6% five-shot Official leaderboard
49 Llama-3.2-3B-Instruct 50.5% five-shot Official leaderboard
50 mT0-xl 50% zero-shot Official leaderboard
51 Llama 3.2 3B 48.4% five-shot Sep 2024 Official leaderboard
52 Qwen2.5-3B 47.5% five-shot Sep 2024 Official leaderboard
53 Qwen2.5-1.5B-Instruct 46.5% five-shot Official leaderboard
54 Gemma 3 1B 45.7% five-shot Mar 2025 Official leaderboard
55 Qwen2.5-1.5B 42.9% five-shot Sep 2024 Official leaderboard
56 Qwen3-0.6B 41% five-shot Apr 2025 Official leaderboard
57 Llama-3.2-1B-Instruct 40.5% five-shot Official leaderboard
58 mT0-large 37% zero-shot Official leaderboard
59 Llama-2-7b-chat-hf 36.8% five-shot Official leaderboard
60 Llama 3.2 1B 36.4% zero-shot Sep 2024 Official leaderboard
61 Llama 2-7B 35.3% zero-shot Jul 2023 Official leaderboard
62 BLOOMZ-1b1 35.1% zero-shot Official leaderboard
63 Qwen2.5-0.5B-Instruct 34.9% zero-shot Official leaderboard
64 Qwen2.5-0.5B 34.7% zero-shot Sep 2024 Official leaderboard
65 BLOOMZ-1b7 34.7% zero-shot Official leaderboard
66 XGLM-4.5B 33% zero-shot Official leaderboard
67 XGLM-1.7B 33% zero-shot Official leaderboard
68 XGLM-2.9B 32.7% zero-shot Official leaderboard
69 BLOOMZ-7b1 32.4% zero-shot Official leaderboard
70 EuroLLM-1.7B-Instruct 32.4% five-shot Official leaderboard
71 EuroLLM-1.7B 32.3% zero-shot Official leaderboard
72 XGLM-7.5B 32.2% zero-shot Official leaderboard
73 Gemma 3 1B PT 31.9% zero-shot Mar 2025 Official leaderboard
#ModelScoreSettingDateSource
1 Gemini 3 Flash 93.2% zero-shot Dec 2025 Official leaderboard
2 GPT-5.2 89.2% five-shot Dec 2025 Official leaderboard
3 Gemma 3 27B PT 82.6% five-shot Mar 2025 Official leaderboard
4 Qwen2.5-72B 82.2% five-shot Sep 2024 Official leaderboard
5 Llama 3.3 70B Instruct 81.5% five-shot Dec 2024 Official leaderboard
6 Gemma 3 27B 81.3% five-shot Mar 2025 Official leaderboard
7 Llama 3.1-70B 80.4% five-shot Jul 2024 Official leaderboard
8 Gemma 3 12B PT 77.3% five-shot Mar 2025 Official leaderboard
9 Qwen3-14B 77.3% five-shot Apr 2025 Official leaderboard
10 Gemma 3 12B 76.3% five-shot Mar 2025 Official leaderboard
11 Qwen2.5-32B 75.7% five-shot Sep 2024 Official leaderboard
12 Qwen3-8B 72.8% five-shot Apr 2025 Official leaderboard
13 Qwen2.5-14B 71.1% five-shot Sep 2024 Official leaderboard
14 Mistral Small 3 70.5% zero-shot Jan 2025 Official leaderboard
15 GLM-4-9B 70.2% five-shot Jun 2024 Official leaderboard
16 GLM-4-9B-Chat 68.8% five-shot Jun 2024 Official leaderboard
17 qwen3-4b-instruct-2507 68.3% five-shot Aug 2025 Official leaderboard
18 Qwen3-4B 67.1% five-shot Apr 2025 Official leaderboard
19 Aya Expanse 8B 65.6% five-shot Oct 2024 Official leaderboard
20 Gemma 3 4B 64.8% five-shot Mar 2025 Official leaderboard
21 Gemma 3 4B PT 64.5% five-shot Mar 2025 Official leaderboard
22 Llama 3-8B 64.2% five-shot Apr 2024 Official leaderboard
23 Llama 3.1-8B 63.3% five-shot Jul 2024 Official leaderboard
24 Qwen2.5-7B 63% five-shot Sep 2024 Official leaderboard
25 Qwen3-1.7B 58% five-shot Apr 2025 Official leaderboard
26 Llama 3.2 3B 48.4% five-shot Sep 2024 Official leaderboard
27 Qwen2.5-3B 47.5% five-shot Sep 2024 Official leaderboard
28 Gemma 3 1B 45.7% five-shot Mar 2025 Official leaderboard
29 Qwen2.5-1.5B 42.9% five-shot Sep 2024 Official leaderboard
30 Qwen3-0.6B 41% five-shot Apr 2025 Official leaderboard
31 Llama 3.2 1B 36.4% zero-shot Sep 2024 Official leaderboard
32 Llama 2-7B 35.3% zero-shot Jul 2023 Official leaderboard
33 Qwen2.5-0.5B 34.7% zero-shot Sep 2024 Official leaderboard
34 Gemma 3 1B PT 31.9% zero-shot Mar 2025 Official leaderboard

Rankings by part

We read 5 parts from the benchmark’s paper, each ranking models on one part of the test, such as STEM, Humanities and Social Sciences. Gemini 3 Flash leads every one of them. Arrows show how far a model moved from its overall rank among the models we track.

STEM

34 models we track have a score here.

  1. 1 Gemini 3 Flash 92.8%
  2. 2 GPT-5.2 88.3%
  3. 3 Gemma 3 27B PT 82.3%
  4. 4 Qwen2.5-72B 82.2%
  5. 5 Gemma 3 27B ↑1 80.5%
  6. 6 Llama 3.3 70B Instruct ↓1 79.7%
  7. 7 Qwen3-14B ↑2 78.3%
  8. 8 Llama 3.1-70B ↓1 77.6%
  9. 9 Gemma 3 12B PT ↓1 77.2%
  10. 10 Qwen2.5-32B ↑1 75.6%

Scores at each effort level

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
Gemini 3 Flash
five-shot
93.2%
GPT-5.2
zero-shot
89.2%
Gemma 3 27B PT
zero-shot five-shot
82.6%
Qwen2.5-72B
zero-shot five-shot
82.2%
Llama 3.3 70B Instruct
zero-shot
81.5%
Gemma 3 27B
zero-shot
81.3%
Llama 3.1-70B
zero-shot five-shot
80.4%
Gemma 3 12B PT
zero-shot
77.3%
Qwen3-14B
zero-shot five-shot
77.3%
Gemma 3 12B
zero-shot
76.3%
Qwen2.5-32B
zero-shot
75.7%
Qwen3-8B
zero-shot five-shot
72.8%

Source: The GreekMMLU paper

ModelEffortHarnessScoreCost per taskOutput tokens
Gemini 3 Flash zero-shot 93.2%
Gemini 3 Flash five-shot 91.2%
GPT-5.2 five-shot 89.2%
GPT-5.2 zero-shot 87.8%
Gemma 3 27B PT five-shot 82.6%
Qwen2.5-72B five-shot 82.2%
Llama 3.3 70B Instruct five-shot 81.5%
Qwen2.5-72B five-shot 81.4%
Gemma 3 27B five-shot 81.3%
Llama 3.1-70B five-shot 80.4%
Qwen2.5-72B zero-shot 79.7%
Llama 3.3 70B Instruct zero-shot 79.7%
Gemma 3 27B zero-shot 79.4%
Qwen2.5-72B zero-shot 79.2%
Gemma 3 27B PT zero-shot 78.9%
Gemma 3 12B PT five-shot 77.3%
Qwen3-14B five-shot 77.3%
Gemma 3 12B five-shot 76.3%
Qwen2.5-32B five-shot 75.7%
Llama 3.1-70B zero-shot 75.7%
Gemma 3 12B zero-shot 75.3%
Gemma 3 12B PT zero-shot 75%
Qwen2.5-32B zero-shot 73.1%
Qwen3-8B five-shot 72.8%
Qwen2.5-14B five-shot 71.1%
Mistral Small 3 zero-shot 70.5%
GLM-4-9B five-shot 70.2%
GLM-4-9B-Chat five-shot 68.8%
Qwen3-8B zero-shot 68.8%
qwen3-4b-instruct-2507 five-shot 68.3%
Qwen3-4B five-shot 67.1%
Aya Expanse 8B five-shot 65.6%
Qwen3-14B zero-shot 65.3%
GLM-4-9B zero-shot 65%
Gemma 3 4B five-shot 64.8%
GLM-4-9B-Chat zero-shot 64.7%
Gemma 3 4B PT five-shot 64.5%
qwen3-4b-instruct-2507 zero-shot 64.5%
Qwen2.5-14B zero-shot 64.4%
Llama 3-8B five-shot 64.2%
Aya Expanse 8B zero-shot 64.2%
Qwen3-4B zero-shot 63.4%
Llama 3.1-8B five-shot 63.3%
Qwen2.5-7B five-shot 63%
Gemma 3 4B zero-shot 62.2%
Qwen3-1.7B five-shot 58%
Qwen2.5-7B zero-shot 55.2%
Llama 3-8B zero-shot 54.1%
Llama 3.1-8B zero-shot 53.1%
Gemma 3 4B PT zero-shot 52.2%
Qwen3-1.7B zero-shot 48.9%
Llama 3.2 3B five-shot 48.4%
Qwen2.5-3B five-shot 47.5%
Gemma 3 1B five-shot 45.7%
Gemma 3 1B zero-shot 45%
Qwen2.5-3B zero-shot 44.9%
Qwen2.5-1.5B five-shot 42.9%
Llama 3.2 3B zero-shot 42.8%
Qwen3-0.6B five-shot 41%
Llama 3.2 1B zero-shot 36.4%
Qwen2.5-1.5B zero-shot 35.9%
Qwen3-0.6B zero-shot 35.7%
Llama 2-7B zero-shot 35.3%
Llama 2-7B five-shot 35%
Qwen2.5-0.5B zero-shot 34.7%
Llama 3.2 1B five-shot 34.6%
Qwen2.5-0.5B five-shot 33.4%
Gemma 3 1B PT zero-shot 31.9%
Gemma 3 1B PT five-shot 30.8%

Scores at each effort level

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
Gemini 3 Flash
five-shot
93.2%
GPT-5.2
zero-shot
89.2%
Gemma 3 27B PT
zero-shot five-shot
82.6%
Qwen2.5-72B
zero-shot five-shot
82.2%
Llama 3.3 70B Instruct
zero-shot
81.5%
Gemma 3 27B
zero-shot
81.3%
Llama 3.1-70B
zero-shot five-shot
80.4%
Gemma 3 12B PT
zero-shot
77.3%
Qwen3-14B
zero-shot five-shot
77.3%
Gemma 3 12B
zero-shot
76.3%
Qwen2.5-32B
zero-shot
75.7%
Qwen3-8B
zero-shot five-shot
72.8%

Source: The GreekMMLU paper

ModelEffortHarnessScoreCost per taskOutput tokens
Gemini 3 Flash zero-shot 93.2%
Gemini 3 Flash five-shot 91.2%
GPT-5.2 five-shot 89.2%
GPT-5.2 zero-shot 87.8%
Gemma 3 27B PT five-shot 82.6%
Qwen2.5-72B five-shot 82.2%
Llama 3.3 70B Instruct five-shot 81.5%
Qwen2.5-72B five-shot 81.4%
Gemma 3 27B five-shot 81.3%
Llama 3.1-70B five-shot 80.4%
Qwen2.5-72B zero-shot 79.7%
Llama 3.3 70B Instruct zero-shot 79.7%
Gemma 3 27B zero-shot 79.4%
Qwen2.5-72B zero-shot 79.2%
Gemma 3 27B PT zero-shot 78.9%
Gemma 3 12B PT five-shot 77.3%
Qwen3-14B five-shot 77.3%
Gemma 3 12B five-shot 76.3%
Qwen2.5-32B five-shot 75.7%
Llama 3.1-70B zero-shot 75.7%
Gemma 3 12B zero-shot 75.3%
Gemma 3 12B PT zero-shot 75%
Qwen2.5-32B zero-shot 73.1%
Qwen3-8B five-shot 72.8%
Qwen2.5-14B five-shot 71.1%
Mistral Small 3 zero-shot 70.5%
GLM-4-9B five-shot 70.2%
GLM-4-9B-Chat five-shot 68.8%
Qwen3-8B zero-shot 68.8%
qwen3-4b-instruct-2507 five-shot 68.3%
Qwen3-4B five-shot 67.1%
Aya Expanse 8B five-shot 65.6%
Qwen3-14B zero-shot 65.3%
GLM-4-9B zero-shot 65%
Gemma 3 4B five-shot 64.8%
GLM-4-9B-Chat zero-shot 64.7%
Gemma 3 4B PT five-shot 64.5%
qwen3-4b-instruct-2507 zero-shot 64.5%
Qwen2.5-14B zero-shot 64.4%
Llama 3-8B five-shot 64.2%
Aya Expanse 8B zero-shot 64.2%
Qwen3-4B zero-shot 63.4%
Llama 3.1-8B five-shot 63.3%
Qwen2.5-7B five-shot 63%
Gemma 3 4B zero-shot 62.2%
Qwen3-1.7B five-shot 58%
Qwen2.5-7B zero-shot 55.2%
Llama 3-8B zero-shot 54.1%
Llama 3.1-8B zero-shot 53.1%
Gemma 3 4B PT zero-shot 52.2%
Qwen3-1.7B zero-shot 48.9%
Llama 3.2 3B five-shot 48.4%
Qwen2.5-3B five-shot 47.5%
Gemma 3 1B five-shot 45.7%
Gemma 3 1B zero-shot 45%
Qwen2.5-3B zero-shot 44.9%
Qwen2.5-1.5B five-shot 42.9%
Llama 3.2 3B zero-shot 42.8%
Qwen3-0.6B five-shot 41%
Llama 3.2 1B zero-shot 36.4%
Qwen2.5-1.5B zero-shot 35.9%
Qwen3-0.6B zero-shot 35.7%
Llama 2-7B zero-shot 35.3%
Llama 2-7B five-shot 35%
Qwen2.5-0.5B zero-shot 34.7%
Llama 3.2 1B five-shot 34.6%
Qwen2.5-0.5B five-shot 33.4%
Gemma 3 1B PT zero-shot 31.9%
Gemma 3 1B PT five-shot 30.8%

Number one over time

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

Google DeepMind
Benchmark paper Passed 85% in Feb 2026, by Gemini 3 Flash

Source: The GreekMMLU paper

GPT-5.2 on the official leaderboard passed 85% in December 2025

Models released before it came out already scored that high. All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
In the benchmark’s paper Best so far
0% 20% 40% 60% 80% 100% 2024 2025 2026 Near the ceiling (85%) Gemini 3 Flash Gemma 3 27B PT Qwen2.5-72B GLM-4-9B Llama 3-8B Llama 2-7B 0% 20% 40% 60% 80% 100% 2024 2025 2026 Near the ceiling (85%) Gemini 3 Flash Gemma 3 27B PT

Source: Official GreekMMLU 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 8 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 2 pass 3 notes
  • Pass Independent runs Scores from the benchmark’s paper (73 entries).
  • Pass Models scored 34 models we track, plus 39 other entries on its leaderboard.
  • Note Newest score The newest came 11 months ago, in December 2025, for Gemini 3 Flash. That’s expected for a saturated test.
  • Note Frontier models None of the top 10 has a score, which is fine for a retired test. The strongest tested is GPT-5.2 (#51).
  • Note Headroom Saturated. The best is 93.2%, past 85% since December 2025.
Leaderboard import 3 pass
  • Pass Importer We read the board on arxiv.org: 69 rows for models we track and 39 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