Benchmark Almanac

Open Arabic LLM Leaderboard v2

Applied-Innovation-Center Karnak 79.3% tops its leaderboard Qwen2.5-72B 72.4% leads the models we track Dropped

Models are scored on native Arabic tasks (AlGhafa native subsets, ArabicMMLU, Exams, MadinahQA, AraTrust, ALRAGE, ArbMMLU-HT) and ranked by the average.

Version 2 dropped saturated and machine-translated tasks from version 1 (2024) and fixed a scoring bug in AlGhafa, so scores are not comparable with version 1. ALRAGE is judged by a model.

Made by Open Arabic LLM Leaderboard team (2A2I and community) with Hugging Face Released Feb 2025 Knowledge in many languages All tasks public Last in a launch post May 2025 , dropped before it saturated Last model added Nov 2025 Scores refreshed 2 Nov 2026
Level
School Students, from grade school to high school
Scored
Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
Answers
Multiple choice
Leaderboard huggingface.co
151
Entries on its own leaderboard, 39 models we track
1
Launch post lists it, from 1 lab
1
System card reports it, from 1 lab

Applied-Innovation-Center Karnak tops the leaderboard with 79.3%

Best score per entry from the benchmark's own leaderboard, read on 2 Nov 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. The best model we track is Qwen2.5-72B, at 72.4%.

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

151 entries. Scroll sideways for the rest.

  1. 79.3 Applied-Innovation-Center Karnak
  2. 75.9 deep-analysis-research D2IL-Arabic-Qwen2.5-72B-Instruct-v0.1
  3. 75.8 deep-analysis-research D2IL-Arabic-Qwen2.5-72B-Instruct-v0.2
  4. 75.6 rombodawg Rombos-LLM-V2.5-Qwen-72b
  5. 75.6 Lina-Z qwen_arabic_ft
  6. 75.5 Sakalti Ultiima-72B
  7. 75.5 AbdulmalekDS qwen72b-ar-lora
  8. 75.5 tunny Arabic_Qwen2.5_72B_instruct_finetune_0.1
  9. 74.8 Lina-Z arabic_llm_test
  10. 73.9 Applied-Innovation-Center AIC-1
  11. 73.1 MaziyarPanahi calme-2.1-qwen2.5-72b
  12. 72.7 MaziyarPanahi calme-2.2-qwen2.5-72b
  13. 72.6 Mushari440 Qwen3-8B-SFT-V2
  14. 72.4 Qwen2.5-72B
  15. 72.0 maldv Qwentile2.5-32B-Instruct
  16. 71.9 Sakalti Saka-14B
  17. 71.9 Sakalti SakaMoe-3x14B-Instruct
  18. 71.6 Cran-May tempmotacilla-cinerea-0308
  19. 71.4 Gemma 3 27B
  20. 71.1 hammh0a Hala-9B
  21. 70.9 FreedomIntelligence AceGPT-v2-32B-Chat
  22. 70.8 tanliboy lambda-qwen2.5-32b-dpo-test
  23. 70.7 freewheelin free-evo-qwen72b-v0.8-re
  24. 70.4 maldv Awqward2.5-32B-Instruct
  25. 70.3 QCRI/Fanar-1-9B-Instruct
  26. 70.3 zetasepic Qwen2.5-32B-Instruct-abliterated-v2
  27. 70.1 Qwen2-72B
  28. 70.1 TIGER-Lab Qwen2.5-32B-Instruct-CFT
  29. 70.1 FreedomIntelligence AceGPT-v2-70B-Chat
  30. 70.0 Qwen2.5-32B
  31. 69.9 TarjamaN Pronoia-14b-community
  32. 69.8 Daemontatox Cogito-R1
  33. 69.5 huihui-ai Qwen2.5-32B-Instruct-abliterated
  34. 69.3 Llama 3.3 70B Instruct
  35. 69.2 Mushari440 qwen3-8B-SFT
  36. 68.5 Qwen2.5-14B
  37. 68.3 tanliboy lambda-qwen2.5-14b-dpo-test
  38. 68.0 AXCXEPT EZO-Qwen2.5-32B-Instruct
  39. 67.2 FreedomIntelligence AceGPT-v2-70B
  40. 67.2 Aya Expanse 32B
  41. 67.1 CohereForAI/c4ai-command-r7b-arabic-02-2025
  42. 66.6 Mushari440 qwen-3-8B-sft-grpo
  43. 66.4 recoilme recoilme-gemma-2-9B-v0.4
  44. 66.2 Qwen3-8B
  45. 66.1 CohereForAI/aya-23-35B
  46. 65.7 JasperV13 Yehia-7B-DPO-Reasoning-preview
  47. 65.7 Navid-AI Yehia-7B-preview
  48. 65.5 JasperV13 Yehia-7B-Reasoning-preview
  49. 65.4 inceptionai/jais-family-30b-16k-chat
  50. 65.3 inceptionai/jais-adapted-70b-chat
  51. 65.3 ALLaM-AI/ALLaM-7B-Instruct-preview
  52. 64.9 inceptionai/jais-family-30b-8k-chat
  53. 64.1 v000000 Qwen2.5-Lumen-14B
  54. 63.9 rombodawg Rombos-LLM-V2.6-Qwen-14b
  55. 63.8 Ocelotr Qwen-8B-SFT
  56. 63.6 Qwen2-7B-Instruct
  57. 63.5 v000000 Qwen2.5-14B-Gutenberg-1e-Delta
  58. 63.5 Ocelotr Qwen3-8B-GAE
  59. 63.0 Cran-May T.E-8.1
  60. 62.9 Qwen3-4B
  61. 62.8 QCRI/Fanar-1-9B
  62. 62.8 amine-khelif BARS
  63. 62.5 huihui-ai Qwen2.5-7B-Instruct-abliterated-v2
  64. 62.4 JasperV13 Yehia-7B-SFT-Reasoning-preview
  65. 62.4 FreedomIntelligence AceGPT-v2-8B-Chat
  66. 62.3 arcee-ai Arcee-Spark
  67. 61.9 SeaLLMs SeaLLM-7B-v2.5
  68. 61.8 Orion-zhen Qwen2.5-7B-Instruct-Uncensored
  69. 61.8 asas-ai AceGPT-v2-32B
  70. 61.7 FreedomIntelligence AceGPT-v2-32B
  71. 61.1 Qwen2.5-7B
  72. 61.1 AIDC-AI Marco-LLM-AR-V4
  73. 60.9 Aya Expanse 8B
  74. 59.7 Isaak-Carter Josiefied-Qwen2.5-7B-Instruct-abliterated-v2
  75. 58.7 airev-ai emirati-14b-v2
  76. 58.7 recoilme recoilme-gemma-2-9B-v0.2
  77. 58.6 AIDC-AI Marco-LLM-AR-V2
  78. 58.2 Gemma 3 12B
  79. 58.1 SeaLLMs SeaLLMs-v3-7B-Chat
  80. 58.1 inceptionai/jais-adapted-13b-chat
  81. 57.7 Slim205 Barka-9b-it
  82. 57.6 silma-ai SILMA-9B-Instruct-v1.0
  83. 57.6 VAGOsolutions SauerkrautLM-Nemo-12b-Instruct
  84. 57.0 huihui-ai Qwen2.5-7B-Instruct-abliterated
  85. 56.2 Qwen1.5-32B
  86. 56.0 Sakalti oxyge1-33B
  87. 55.8 Phi-4 Mini
  88. 55.5 Gemma 2 27B
  89. 55.4 Llama 3.1-8B
  90. 55.0 FreedomIntelligence AceGPT-v1.5-13B-Chat
  91. 54.2 Qwen2.5-3B
  92. 53.8 Qwen3-14B
  93. 53.6 inceptionai/jais-family-30b-8k
  94. 52.9 Qwen3-1.7B
  95. 52.7 01-ZeroOne SUHAIL-14B-preview
  96. 52.5 FreedomIntelligence AceGPT-13B-chat
  97. 52.5 Qwen1.5-14B
  98. 52.4 silma-ai SILMA-Kashif-2B-Instruct-v1.0
  99. 52.4 inceptionai/jais-family-30b-16k
  100. 52.1 inceptionai/jais-family-13b-chat
  101. 51.9 inceptionai/jais-adapted-70b
  102. 51.0 inceptionai/jais-adapted-7b-chat
  103. 50.5 Gemma 3 4B
  104. 50.5 Hagrass LLama3-3.2-instruct-trained
  105. 49.5 upstage SOLAR-10.7B-v1.0
  106. 49.2 Qwen2-1.5B-Instruct
  107. 49.1 FreedomIntelligence AceGPT-7B-chat
  108. 48.6 Qwen3-30B-A3B-Instruct
  109. 47.3 yellowtown 7B-v0.2
  110. 47.2 Qwen1.5-7B
  111. 47.2 FreedomIntelligence AceGPT-13B
  112. 47.2 Qwen2.5-1.5B
  113. 46.6 Gemma 2 9B
  114. 46.5 AIDC-AI Marco-LLM-AR
  115. 45.7 Slim205 Barka-2b-it
  116. 45.7 Phi-4
  117. 44.9 01-ai/Yi-1.5-9B
  118. 43.7 FreedomIntelligence AceGPT-v1.5-13B
  119. 42.6 INSAIT-Institute BgGPT-7B-Instruct-v0.2
  120. 42.5 inceptionai/jais-adapted-13b
  121. 42.1 speakleash Bielik-11B-v2
  122. 41.8 cognitivecomputations Dolphin3.0-R1-Mistral-24B
  123. 41.2 Qwen3-32B
  124. 41.1 tiiuae/Falcon3-10B-Base
  125. 41.1 Llama 3.2 3B
  126. 40.1 Syed-Hasan-8503 Phi-3-mini-4K-instruct-cpo-simpo
  127. 39.6 01-ai/Yi-1.5-9B-32K
  128. 39.1 Sakalti Saka-1.5B
  129. 38.8 Qwen/Qwen1.5-4B
  130. 38.7 Naseej noon-7b
  131. 38.6 inceptionai/jais-family-13b
  132. 38.5 01-ai/Yi-1.5-34B-32K
  133. 38.3 AIDC-AI Marco-LLM-AR-V3
  134. 38.0 Qwen3-0.6B
  135. 36.9 qwen3-4b-instruct-2507
  136. 36.8 FreedomIntelligence AceGPT-7B
  137. 36.5 tiiuae/Falcon3-7B-Base
  138. 35.6 Qwen2-0.5B-Instruct
  139. 35.1 DeepSeek-R1-Distill-Qwen-32B
  140. 34.7 Gemma 2 2B IT
  141. 34.4 riotu-lab ArabianGPT-01B
  142. 34.2 Qwen2.5-0.5B
  143. 32.8 QwQ-32B
  144. 32.4 DeepSeek-R1-Distill-Qwen-14B
  145. 31.5 mobiuslabsgmbh DeepSeek-R1-ReDistill-Llama3-8B-v1.1
  146. 30.6 inceptionai/jais-adapted-7b
  147. 30.2 Llama 3.2 1B
  148. 29.9 TIGER-Lab Qwen2.5-Math-7B-CFT
  149. 29.5 Qwen/Qwen1.5-1.8B
  150. 28.4 mobiuslabsgmbh DeepSeek-R1-ReDistill-Qwen-7B-v1.1
  151. 27.7 stabilityai ar-stablelm-2-base
  1. 1 Applied-Innovation-Center Karnak 79.3%
  2. 2 deep-analysis-research D2IL-Arabic-Qwen2.5-72B-Instruct-v0.1 75.9%
  3. 3 deep-analysis-research D2IL-Arabic-Qwen2.5-72B-Instruct-v0.2 75.8%
  4. 4 rombodawg Rombos-LLM-V2.5-Qwen-72b 75.6%
  5. 5 Lina-Z qwen_arabic_ft 75.6%
  6. 6 Sakalti Ultiima-72B 75.5%
  7. 7 AbdulmalekDS qwen72b-ar-lora 75.5%
  8. 8 tunny Arabic_Qwen2.5_72B_instruct_finetune_0.1 75.5%
  9. 9 Lina-Z arabic_llm_test 74.8%
  10. 10 Applied-Innovation-Center AIC-1 73.9%
  11. 11 MaziyarPanahi calme-2.1-qwen2.5-72b 73.1%
  12. 12 MaziyarPanahi calme-2.2-qwen2.5-72b 72.7%
  13. 13 Mushari440 Qwen3-8B-SFT-V2 72.6%
  14. 14 Qwen2.5-72B 72.4%
  15. 15 maldv Qwentile2.5-32B-Instruct 72%
  16. 16 Sakalti Saka-14B 71.9%
  17. 17 Sakalti SakaMoe-3x14B-Instruct 71.9%
  18. 18 Cran-May tempmotacilla-cinerea-0308 71.6%
  19. 19 Gemma 3 27B 71.4%
  20. 20 hammh0a Hala-9B 71.1%
  21. 21 FreedomIntelligence AceGPT-v2-32B-Chat 70.9%
  22. 22 tanliboy lambda-qwen2.5-32b-dpo-test 70.8%
  23. 23 freewheelin free-evo-qwen72b-v0.8-re 70.7%
  24. 24 maldv Awqward2.5-32B-Instruct 70.4%
  25. 25 QCRI/Fanar-1-9B-Instruct 70.3%
  26. 26 zetasepic Qwen2.5-32B-Instruct-abliterated-v2 70.3%
  27. 27 Qwen2-72B base model 70.1%
  28. 28 TIGER-Lab Qwen2.5-32B-Instruct-CFT 70.1%
  29. 29 FreedomIntelligence AceGPT-v2-70B-Chat 70.1%
  30. 30 Qwen2.5-32B 70%
  31. 31 TarjamaN Pronoia-14b-community 69.9%
  32. 32 Daemontatox Cogito-R1 69.8%
  33. 33 huihui-ai Qwen2.5-32B-Instruct-abliterated 69.5%
  34. 34 Llama 3.3 70B Instruct 69.3%
  35. 35 Mushari440 qwen3-8B-SFT 69.2%
  36. 36 Qwen2.5-14B base model 68.5%
  37. 37 tanliboy lambda-qwen2.5-14b-dpo-test 68.3%
  38. 38 AXCXEPT EZO-Qwen2.5-32B-Instruct 68%
  39. 39 FreedomIntelligence AceGPT-v2-70B 67.2%
  40. 40 Aya Expanse 32B 67.2%
  41. 41 CohereForAI/c4ai-command-r7b-arabic-02-2025 67.1%
  42. 42 Mushari440 qwen-3-8B-sft-grpo 66.6%
  43. 43 recoilme recoilme-gemma-2-9B-v0.4 66.4%
  44. 44 Qwen3-8B base model 66.2%
  45. 45 CohereForAI/aya-23-35B 66.1%
  46. 46 JasperV13 Yehia-7B-DPO-Reasoning-preview 65.7%
  47. 47 Navid-AI Yehia-7B-preview 65.7%
  48. 48 JasperV13 Yehia-7B-Reasoning-preview 65.5%
  49. 49 inceptionai/jais-family-30b-16k-chat 65.4%
  50. 50 inceptionai/jais-adapted-70b-chat 65.3%
  51. 51 ALLaM-AI/ALLaM-7B-Instruct-preview 65.3%
  52. 52 inceptionai/jais-family-30b-8k-chat 64.9%
  53. 53 v000000 Qwen2.5-Lumen-14B 64.1%
  54. 54 rombodawg Rombos-LLM-V2.6-Qwen-14b 63.9%
  55. 55 Ocelotr Qwen-8B-SFT 63.8%
  56. 56 Qwen2-7B-Instruct 63.6%
  57. 57 v000000 Qwen2.5-14B-Gutenberg-1e-Delta 63.5%
  58. 58 Ocelotr Qwen3-8B-GAE 63.5%
  59. 59 Cran-May T.E-8.1 63%
  60. 60 Qwen3-4B base model 62.9%
  61. 61 QCRI/Fanar-1-9B 62.8%
  62. 62 amine-khelif BARS 62.8%
  63. 63 huihui-ai Qwen2.5-7B-Instruct-abliterated-v2 62.5%
  64. 64 JasperV13 Yehia-7B-SFT-Reasoning-preview 62.4%
  65. 65 FreedomIntelligence AceGPT-v2-8B-Chat 62.4%
  66. 66 arcee-ai Arcee-Spark 62.3%
  67. 67 SeaLLMs SeaLLM-7B-v2.5 61.9%
  68. 68 Orion-zhen Qwen2.5-7B-Instruct-Uncensored 61.8%
  69. 69 asas-ai AceGPT-v2-32B 61.8%
  70. 70 FreedomIntelligence AceGPT-v2-32B 61.7%
  71. 71 Qwen2.5-7B base model 61.1%
  72. 72 AIDC-AI Marco-LLM-AR-V4 61.1%
  73. 73 Aya Expanse 8B 60.9%
  74. 74 Isaak-Carter Josiefied-Qwen2.5-7B-Instruct-abliterated-v2 59.7%
  75. 75 airev-ai emirati-14b-v2 58.7%
  76. 76 recoilme recoilme-gemma-2-9B-v0.2 58.7%
  77. 77 AIDC-AI Marco-LLM-AR-V2 58.6%
  78. 78 Gemma 3 12B base model 58.2%
  79. 79 SeaLLMs SeaLLMs-v3-7B-Chat 58.1%
  80. 80 inceptionai/jais-adapted-13b-chat 58.1%
  81. 81 Slim205 Barka-9b-it 57.7%
  82. 82 silma-ai SILMA-9B-Instruct-v1.0 57.6%
  83. 83 VAGOsolutions SauerkrautLM-Nemo-12b-Instruct 57.6%
  84. 84 huihui-ai Qwen2.5-7B-Instruct-abliterated 57%
  85. 85 Qwen1.5-32B base model 56.2%
  86. 86 Sakalti oxyge1-33B 56%
  87. 87 Phi-4 Mini 55.8%
  88. 88 Gemma 2 27B 55.5%
  89. 89 Llama 3.1-8B 55.4%
  90. 90 FreedomIntelligence AceGPT-v1.5-13B-Chat 55%
  91. 91 Qwen2.5-3B 54.2%
  92. 92 Qwen3-14B base model 53.8%
  93. 93 inceptionai/jais-family-30b-8k 53.6%
  94. 94 Qwen3-1.7B base model 52.9%
  95. 95 01-ZeroOne SUHAIL-14B-preview 52.7%
  96. 96 FreedomIntelligence AceGPT-13B-chat 52.5%
  97. 97 Qwen1.5-14B base model 52.5%
  98. 98 silma-ai SILMA-Kashif-2B-Instruct-v1.0 52.4%
  99. 99 inceptionai/jais-family-30b-16k 52.4%
  100. 100 inceptionai/jais-family-13b-chat 52.1%
  101. 101 inceptionai/jais-adapted-70b 51.9%
  102. 102 inceptionai/jais-adapted-7b-chat 51%
  103. 103 Gemma 3 4B base model 50.5%
  104. 104 Hagrass LLama3-3.2-instruct-trained 50.5%
  105. 105 upstage SOLAR-10.7B-v1.0 49.5%
  106. 106 Qwen2-1.5B-Instruct 49.2%
  107. 107 FreedomIntelligence AceGPT-7B-chat 49.1%
  108. 108 Qwen3-30B-A3B-Instruct Jul 2025 48.6%
  109. 109 yellowtown 7B-v0.2 47.3%
  110. 110 Qwen1.5-7B base model 47.2%
  111. 111 FreedomIntelligence AceGPT-13B 47.2%
  112. 112 Qwen2.5-1.5B 47.2%
  113. 113 Gemma 2 9B 46.6%
  114. 114 AIDC-AI Marco-LLM-AR 46.5%
  115. 115 Slim205 Barka-2b-it 45.7%
  116. 116 Phi-4 45.7%
  117. 117 01-ai/Yi-1.5-9B base model 44.9%
  118. 118 FreedomIntelligence AceGPT-v1.5-13B 43.7%
  119. 119 INSAIT-Institute BgGPT-7B-Instruct-v0.2 42.6%
  120. 120 inceptionai/jais-adapted-13b 42.5%
  121. 121 speakleash Bielik-11B-v2 42.1%
  122. 122 cognitivecomputations Dolphin3.0-R1-Mistral-24B 41.8%
  123. 123 Qwen3-32B 41.2%
  124. 124 tiiuae/Falcon3-10B-Base base model 41.1%
  125. 125 Llama 3.2 3B 41.1%
  126. 126 Syed-Hasan-8503 Phi-3-mini-4K-instruct-cpo-simpo 40.1%
  127. 127 01-ai/Yi-1.5-9B-32K base model 39.6%
  128. 128 Sakalti Saka-1.5B 39.1%
  129. 129 Qwen/Qwen1.5-4B base model 38.8%
  130. 130 Naseej noon-7b 38.7%
  131. 131 inceptionai/jais-family-13b 38.6%
  132. 132 01-ai/Yi-1.5-34B-32K base model 38.5%
  133. 133 AIDC-AI Marco-LLM-AR-V3 38.3%
  134. 134 Qwen3-0.6B 38%
  135. 135 qwen3-4b-instruct-2507 36.9%
  136. 136 FreedomIntelligence AceGPT-7B 36.8%
  137. 137 tiiuae/Falcon3-7B-Base base model 36.5%
  138. 138 Qwen2-0.5B-Instruct 35.6%
  139. 139 DeepSeek-R1-Distill-Qwen-32B 35.1%
  140. 140 Gemma 2 2B IT 34.7%
  141. 141 riotu-lab ArabianGPT-01B 34.4%
  142. 142 Qwen2.5-0.5B 34.2%
  143. 143 QwQ-32B 32.8%
  144. 144 DeepSeek-R1-Distill-Qwen-14B 32.4%
  145. 145 mobiuslabsgmbh DeepSeek-R1-ReDistill-Llama3-8B-v1.1 31.5%
  146. 146 inceptionai/jais-adapted-7b 30.6%
  147. 147 Llama 3.2 1B base model 30.2%
  148. 148 TIGER-Lab Qwen2.5-Math-7B-CFT 29.9%
  149. 149 Qwen/Qwen1.5-1.8B base model 29.5%
  150. 150 mobiuslabsgmbh DeepSeek-R1-ReDistill-Qwen-7B-v1.1 28.4%
  151. 151 stabilityai ar-stablelm-2-base 27.7%
#EntryScoreSettingDateSource
1 Applied-Innovation-Center Karnak 79.3% Feb 2026 Official leaderboard
2 deep-analysis-research D2IL-Arabic-Qwen2.5-72B-Instruct-v0.1 75.9% Jul 2025 Official leaderboard
3 deep-analysis-research D2IL-Arabic-Qwen2.5-72B-Instruct-v0.2 75.8% Sep 2025 Official leaderboard
4 rombodawg Rombos-LLM-V2.5-Qwen-72b 75.6% Feb 2025 Official leaderboard
5 Lina-Z qwen_arabic_ft 75.6% Jan 2026 Official leaderboard
6 Sakalti Ultiima-72B 75.5% Feb 2025 Official leaderboard
7 AbdulmalekDS qwen72b-ar-lora 75.5% Sep 2025 Official leaderboard
8 tunny Arabic_Qwen2.5_72B_instruct_finetune_0.1 75.5% Dec 2025 Official leaderboard
9 Lina-Z arabic_llm_test 74.8% Nov 2025 Official leaderboard
10 Applied-Innovation-Center AIC-1 73.9% Sep 2025 Official leaderboard
11 MaziyarPanahi calme-2.1-qwen2.5-72b 73.1% Jan 2025 Official leaderboard
12 MaziyarPanahi calme-2.2-qwen2.5-72b 72.7% Jan 2025 Official leaderboard
13 Mushari440 Qwen3-8B-SFT-V2 72.6% Mar 2026 Official leaderboard
14 Qwen2.5-72B 72.4% Sep 2024 Official leaderboard
15 maldv Qwentile2.5-32B-Instruct 72% Feb 2025 Official leaderboard
16 Sakalti Saka-14B 71.9% Feb 2025 Official leaderboard
17 Sakalti SakaMoe-3x14B-Instruct 71.9% Feb 2025 Official leaderboard
18 Cran-May tempmotacilla-cinerea-0308 71.6% Mar 2025 Official leaderboard
19 Gemma 3 27B 71.4% Mar 2025 Official leaderboard
20 hammh0a Hala-9B 71.1% Feb 2026 Official leaderboard
21 FreedomIntelligence AceGPT-v2-32B-Chat 70.9% Jan 2025 Official leaderboard
22 tanliboy lambda-qwen2.5-32b-dpo-test 70.8% Jan 2025 Official leaderboard
23 freewheelin free-evo-qwen72b-v0.8-re 70.7% Jan 2025 Official leaderboard
24 maldv Awqward2.5-32B-Instruct 70.4% Feb 2025 Official leaderboard
25 QCRI/Fanar-1-9B-Instruct 70.3% Jun 2025 Official leaderboard
26 zetasepic Qwen2.5-32B-Instruct-abliterated-v2 70.3% Mar 2025 Official leaderboard
27 Qwen2-72B 70.1% base model Jun 2024 Official leaderboard
28 TIGER-Lab Qwen2.5-32B-Instruct-CFT 70.1% Mar 2025 Official leaderboard
29 FreedomIntelligence AceGPT-v2-70B-Chat 70.1% Jan 2025 Official leaderboard
30 Qwen2.5-32B 70% Sep 2024 Official leaderboard
31 TarjamaN Pronoia-14b-community 69.9% Feb 2025 Official leaderboard
32 Daemontatox Cogito-R1 69.8% Feb 2025 Official leaderboard
33 huihui-ai Qwen2.5-32B-Instruct-abliterated 69.5% Jan 2025 Official leaderboard
34 Llama 3.3 70B Instruct 69.3% Dec 2024 Official leaderboard
35 Mushari440 qwen3-8B-SFT 69.2% Feb 2026 Official leaderboard
36 Qwen2.5-14B 68.5% base model Sep 2024 Official leaderboard
37 tanliboy lambda-qwen2.5-14b-dpo-test 68.3% Jan 2025 Official leaderboard
38 AXCXEPT EZO-Qwen2.5-32B-Instruct 68% Jan 2025 Official leaderboard
39 FreedomIntelligence AceGPT-v2-70B 67.2% Jan 2025 Official leaderboard
40 Aya Expanse 32B 67.2% Oct 2024 Official leaderboard
41 CohereForAI/c4ai-command-r7b-arabic-02-2025 67.1% Feb 2025 Official leaderboard
42 Mushari440 qwen-3-8B-sft-grpo 66.6% Feb 2026 Official leaderboard
43 recoilme recoilme-gemma-2-9B-v0.4 66.4% Jan 2025 Official leaderboard
44 Qwen3-8B 66.2% base model Apr 2025 Official leaderboard
45 CohereForAI/aya-23-35B 66.1% Jan 2025 Official leaderboard
46 JasperV13 Yehia-7B-DPO-Reasoning-preview 65.7% May 2025 Official leaderboard
47 Navid-AI Yehia-7B-preview 65.7% Mar 2025 Official leaderboard
48 JasperV13 Yehia-7B-Reasoning-preview 65.5% May 2025 Official leaderboard
49 inceptionai/jais-family-30b-16k-chat 65.4% Jan 2025 Official leaderboard
50 inceptionai/jais-adapted-70b-chat 65.3% Jan 2025 Official leaderboard
51 ALLaM-AI/ALLaM-7B-Instruct-preview 65.3% Feb 2025 Official leaderboard
52 inceptionai/jais-family-30b-8k-chat 64.9% Jan 2025 Official leaderboard
53 v000000 Qwen2.5-Lumen-14B 64.1% Jan 2025 Official leaderboard
54 rombodawg Rombos-LLM-V2.6-Qwen-14b 63.9% Jan 2025 Official leaderboard
55 Ocelotr Qwen-8B-SFT 63.8% Dec 2025 Official leaderboard
56 Qwen2-7B-Instruct 63.6% Jun 2024 Official leaderboard
57 v000000 Qwen2.5-14B-Gutenberg-1e-Delta 63.5% Jan 2025 Official leaderboard
58 Ocelotr Qwen3-8B-GAE 63.5% Jan 2026 Official leaderboard
59 Cran-May T.E-8.1 63% Jan 2025 Official leaderboard
60 Qwen3-4B 62.9% base model Apr 2025 Official leaderboard
61 QCRI/Fanar-1-9B 62.8% Jun 2025 Official leaderboard
62 amine-khelif BARS 62.8% Sep 2025 Official leaderboard
63 huihui-ai Qwen2.5-7B-Instruct-abliterated-v2 62.5% Jan 2025 Official leaderboard
64 JasperV13 Yehia-7B-SFT-Reasoning-preview 62.4% May 2025 Official leaderboard
65 FreedomIntelligence AceGPT-v2-8B-Chat 62.4% Jan 2025 Official leaderboard
66 arcee-ai Arcee-Spark 62.3% Jan 2025 Official leaderboard
67 SeaLLMs SeaLLM-7B-v2.5 61.9% Jan 2025 Official leaderboard
68 Orion-zhen Qwen2.5-7B-Instruct-Uncensored 61.8% Jan 2025 Official leaderboard
69 asas-ai AceGPT-v2-32B 61.8% Jan 2025 Official leaderboard
70 FreedomIntelligence AceGPT-v2-32B 61.7% Jan 2025 Official leaderboard
71 Qwen2.5-7B 61.1% base model Sep 2024 Official leaderboard
72 AIDC-AI Marco-LLM-AR-V4 61.1% Mar 2025 Official leaderboard
73 Aya Expanse 8B 60.9% Oct 2024 Official leaderboard
74 Isaak-Carter Josiefied-Qwen2.5-7B-Instruct-abliterated-v2 59.7% Jan 2025 Official leaderboard
75 airev-ai emirati-14b-v2 58.7% Jan 2025 Official leaderboard
76 recoilme recoilme-gemma-2-9B-v0.2 58.7% Jan 2025 Official leaderboard
77 AIDC-AI Marco-LLM-AR-V2 58.6% Mar 2025 Official leaderboard
78 Gemma 3 12B 58.2% base model Mar 2025 Official leaderboard
79 SeaLLMs SeaLLMs-v3-7B-Chat 58.1% Mar 2025 Official leaderboard
80 inceptionai/jais-adapted-13b-chat 58.1% Jan 2025 Official leaderboard
81 Slim205 Barka-9b-it 57.7% Jun 2025 Official leaderboard
82 silma-ai SILMA-9B-Instruct-v1.0 57.6% Jan 2025 Official leaderboard
83 VAGOsolutions SauerkrautLM-Nemo-12b-Instruct 57.6% Jan 2025 Official leaderboard
84 huihui-ai Qwen2.5-7B-Instruct-abliterated 57% Jan 2025 Official leaderboard
85 Qwen1.5-32B 56.2% base model Feb 2024 Official leaderboard
86 Sakalti oxyge1-33B 56% Mar 2025 Official leaderboard
87 Phi-4 Mini 55.8% Feb 2025 Official leaderboard
88 Gemma 2 27B 55.5% Jun 2024 Official leaderboard
89 Llama 3.1-8B 55.4% Jul 2024 Official leaderboard
90 FreedomIntelligence AceGPT-v1.5-13B-Chat 55% Jan 2025 Official leaderboard
91 Qwen2.5-3B 54.2% Sep 2024 Official leaderboard
92 Qwen3-14B 53.8% base model Apr 2025 Official leaderboard
93 inceptionai/jais-family-30b-8k 53.6% Jan 2025 Official leaderboard
94 Qwen3-1.7B 52.9% base model Apr 2025 Official leaderboard
95 01-ZeroOne SUHAIL-14B-preview 52.7% Jul 2025 Official leaderboard
96 FreedomIntelligence AceGPT-13B-chat 52.5% Jan 2025 Official leaderboard
97 Qwen1.5-14B 52.5% base model Feb 2024 Official leaderboard
98 silma-ai SILMA-Kashif-2B-Instruct-v1.0 52.4% Feb 2025 Official leaderboard
99 inceptionai/jais-family-30b-16k 52.4% Jan 2025 Official leaderboard
100 inceptionai/jais-family-13b-chat 52.1% Jan 2025 Official leaderboard
101 inceptionai/jais-adapted-70b 51.9% Jan 2025 Official leaderboard
102 inceptionai/jais-adapted-7b-chat 51% Jan 2025 Official leaderboard
103 Gemma 3 4B 50.5% base model Mar 2025 Official leaderboard
104 Hagrass LLama3-3.2-instruct-trained 50.5% Aug 2025 Official leaderboard
105 upstage SOLAR-10.7B-v1.0 49.5% Jan 2025 Official leaderboard
106 Qwen2-1.5B-Instruct 49.2% Jun 2024 Official leaderboard
107 FreedomIntelligence AceGPT-7B-chat 49.1% Jan 2025 Official leaderboard
108 Qwen3-30B-A3B-Instruct Jul 2025 48.6% Jul 2025 Official leaderboard
109 yellowtown 7B-v0.2 47.3% Jan 2025 Official leaderboard
110 Qwen1.5-7B 47.2% base model Feb 2024 Official leaderboard
111 FreedomIntelligence AceGPT-13B 47.2% Jan 2025 Official leaderboard
112 Qwen2.5-1.5B 47.2% Sep 2024 Official leaderboard
113 Gemma 2 9B 46.6% Jun 2024 Official leaderboard
114 AIDC-AI Marco-LLM-AR 46.5% Feb 2025 Official leaderboard
115 Slim205 Barka-2b-it 45.7% Jun 2025 Official leaderboard
116 Phi-4 45.7% Dec 2024 Official leaderboard
117 01-ai/Yi-1.5-9B 44.9% base model Jan 2025 Official leaderboard
118 FreedomIntelligence AceGPT-v1.5-13B 43.7% Jan 2025 Official leaderboard
119 INSAIT-Institute BgGPT-7B-Instruct-v0.2 42.6% Jan 2025 Official leaderboard
120 inceptionai/jais-adapted-13b 42.5% Jan 2025 Official leaderboard
121 speakleash Bielik-11B-v2 42.1% Jan 2025 Official leaderboard
122 cognitivecomputations Dolphin3.0-R1-Mistral-24B 41.8% Feb 2025 Official leaderboard
123 Qwen3-32B 41.2% Apr 2025 Official leaderboard
124 tiiuae/Falcon3-10B-Base 41.1% base model Feb 2025 Official leaderboard
125 Llama 3.2 3B 41.1% Sep 2024 Official leaderboard
126 Syed-Hasan-8503 Phi-3-mini-4K-instruct-cpo-simpo 40.1% Jan 2025 Official leaderboard
127 01-ai/Yi-1.5-9B-32K 39.6% base model Jan 2025 Official leaderboard
128 Sakalti Saka-1.5B 39.1% Feb 2025 Official leaderboard
129 Qwen/Qwen1.5-4B 38.8% base model Jan 2025 Official leaderboard
130 Naseej noon-7b 38.7% Mar 2025 Official leaderboard
131 inceptionai/jais-family-13b 38.6% Jan 2025 Official leaderboard
132 01-ai/Yi-1.5-34B-32K 38.5% base model Jan 2025 Official leaderboard
133 AIDC-AI Marco-LLM-AR-V3 38.3% Mar 2025 Official leaderboard
134 Qwen3-0.6B 38% Apr 2025 Official leaderboard
135 qwen3-4b-instruct-2507 36.9% Aug 2025 Official leaderboard
136 FreedomIntelligence AceGPT-7B 36.8% Jan 2025 Official leaderboard
137 tiiuae/Falcon3-7B-Base 36.5% base model Feb 2025 Official leaderboard
138 Qwen2-0.5B-Instruct 35.6% Jun 2024 Official leaderboard
139 DeepSeek-R1-Distill-Qwen-32B 35.1% Jan 2025 Official leaderboard
140 Gemma 2 2B IT 34.7% Jul 2024 Official leaderboard
141 riotu-lab ArabianGPT-01B 34.4% Oct 2025 Official leaderboard
142 Qwen2.5-0.5B 34.2% Sep 2024 Official leaderboard
143 QwQ-32B 32.8% Mar 2025 Official leaderboard
144 DeepSeek-R1-Distill-Qwen-14B 32.4% Jan 2025 Official leaderboard
145 mobiuslabsgmbh DeepSeek-R1-ReDistill-Llama3-8B-v1.1 31.5% Feb 2025 Official leaderboard
146 inceptionai/jais-adapted-7b 30.6% Jan 2025 Official leaderboard
147 Llama 3.2 1B 30.2% base model Sep 2024 Official leaderboard
148 TIGER-Lab Qwen2.5-Math-7B-CFT 29.9% Mar 2025 Official leaderboard
149 Qwen/Qwen1.5-1.8B 29.5% base model Jan 2025 Official leaderboard
150 mobiuslabsgmbh DeepSeek-R1-ReDistill-Qwen-7B-v1.1 28.4% Feb 2025 Official leaderboard
151 stabilityai ar-stablelm-2-base 27.7% Jun 2025 Official leaderboard

Labs report scores for 1 model in their cards. We plot every source in the charts below.

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
Qwen2.5-72B
base model
72.4%
Gemma 3 27B
base model
71.4%
Qwen2.5-32B
base model
70%
Qwen2.5-14B
base model
68.5%
Qwen3-8B
base model
66.2%
Qwen3-4B
base model
62.9%
Qwen2.5-7B
base model
61.1%
Gemma 2 27B
base model
55.5%
Llama 3.1-8B
base model
55.4%
Qwen2.5-3B
base model
54.2%
Qwen3-14B
base model
53.8%
Qwen2.5-1.5B
base model
47.2%
ModelEffortHarnessScoreCost per taskOutput tokens
Qwen2.5-72B 72.4%
Gemma 3 27B 71.4%
Qwen2-72B base model 70.1%
Qwen2.5-32B 70%
Qwen2.5-72B base model 69.4%
Llama 3.3 70B Instruct 69.3%
Qwen2.5-14B base model 68.5%
Aya Expanse 32B 67.2%
Qwen3-8B base model 66.2%
Qwen2.5-32B base model 65.5%
Qwen2-7B-Instruct 63.6%
Gemma 3 27B base model 63.2%
Qwen2.5-14B 63.2%
Qwen3-4B base model 62.9%
Qwen2.5-7B base model 61.1%
Aya Expanse 8B 60.9%
Qwen2.5-7B 59.8%
Gemma 3 12B base model 58.2%
Qwen1.5-32B base model 56.2%
Phi-4 Mini 55.8%
Gemma 2 27B 55.5%
Llama 3.1-8B 55.4%
Qwen2.5-3B 54.2%
Qwen3-14B base model 53.8%
Qwen3-1.7B base model 52.9%
Qwen1.5-14B base model 52.5%
Llama 3.1-8B base model 51.6%
Gemma 3 4B base model 50.5%
Qwen2-1.5B-Instruct 49.2%
Qwen3-30B-A3B-Instruct Jul 2025 48.6%
Qwen1.5-7B base model 47.2%
Qwen2.5-1.5B 47.2%
Gemma 2 9B 46.6%
Phi-4 45.7%
Qwen3-8B 42.4%
Qwen3-32B 41.2%
Llama 3.2 3B 41.1%
Qwen3-14B 40.9%
Qwen2.5-3B base model 40.8%
Qwen3-4B 39.9%
Qwen3-0.6B 38%
qwen3-4b-instruct-2507 36.9%
Gemma 2 27B base model 35.6%
Qwen2-0.5B-Instruct 35.6%
DeepSeek-R1-Distill-Qwen-32B 35.1%
Gemma 2 2B IT 34.7%
Qwen2.5-0.5B 34.2%
QwQ-32B 32.8%
DeepSeek-R1-Distill-Qwen-14B 32.4%
Llama 3.2 1B base model 30.2%
Qwen2.5-0.5B base model 26.6%
Qwen2.5-1.5B base model 23.7%

Number one over time

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

Qwen
Own leaderboard Not tested since Aug 2025
Qwen2.5-72B 72.4%

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

On the benchmark’s own leaderboard Reported by the lab in its model card Best so far
0% 20% 40% 60% 80% 100% 2024 2025 2026 Near the ceiling (85%) Qwen2.5-72B Qwen1.5-32B 0% 20% 40% 60% 80% 100% 2024 2025 2026 Near the ceiling (85%) Qwen2.5-72B Qwen1.5-32B

No launch post has listed it since May 2025

Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.

Technology Innovation Institute
DateLaunch postScoreAs the post writes it
21 May 2025 Falcon-Arabic-7B 68 OALL v2 Average

One system card reports it

Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.

DateDocumentScoreAs the document writes itSection
8 Nov 2025 Falcon-Arabic-7B technical report 68.3 OALL benchmark average over 7 tasks, Falcon-Arabic-Instruct (Table 6) Capabilities

Versions

Replaces Open Arabic LLM Leaderboard v1

More in Multilingual

All 115 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. Inkling 88.7%
  • 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%