PerMMLU
Models answer Persian multiple-choice questions from school curricula, university entrance exams, and Iranian cultural sources.
MIZAN ranks 45 model configurations. The test includes 80 cinema questions, 40 poetry questions, and 100 questions about proverbs.
- Made by
- MCINext
- Released
- May 2025
- Kind
- Knowledge in many languages
- Level
- School
- Scored
- Checked automatically
- Answers
- Multiple choice
- Tasks
- All public
- Measures
- Capability
- Subject
- Languages and literature, Arts, Education
- Language
- Persian
- Interaction
- Single turn
- Items from
- Exams and contests, Real work and use
- Refresh
- One fixed set
- Run by
- Anyone
- Strongest model tested
- o3 #84 on the capability index
- Last model added
- September 2025
- Scores refreshed
- November 20th, 2026
Gemini 2.5 Pro leads with 84% among models released by September 2025 Gemini 2.5 Pro leads with 84% among models released by September 2025
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.
The newest model with a score here came out in September 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 84.0
Gemini 2.5 Pro
- 82.2 o3
- 76.2 GPT-5 mini
- 72.8
Gemini 2.5 Flash
- 72.2 Claude 3.7 Sonnet
- 70.4 GPT-4.1
- 69.6
Gemini 2.0 Flash
- 68.8 GPT-4o
- 68.6 GPT-5 nano
- 68.1 deepseek-reasoner
- 67.9 gpt-oss-120b
- 64.7
Gemini 2.0 Flash-Lite
- 61.3 GPT-4.1 mini
- 59.8
Cohere Command A
- 59.6 gpt-oss-20b
- 59.1 deepseek-chat
- 57.1
Llama 3.3 70B Instruct
- 57.0 GPT-4o mini
- 56.4
Llama 4 Scout
- 56.4
Qwen3-32B
- 56.3
Gemma 3 27B
- 54.4 Claude 3.5 Haiku
- 51.4
Qwen3-14B
- 51.0
Gemma 3 12B
- 51.0
Qwen3-30B-A3B
- 49.8 deepseek-r1-0528-qwen3-8b
- 48.0
Command R+
- 47.8 GPT-4.1 nano
- 47.6
Mistral Small 3.1
- 46.8
Gemma 3n E4B
- 44.7
Qwen3-8B
- 41.4
Aya Expanse 32B
- 40.5
Command R (03-2024)
- 40.3
Qwen3-4B
- 38.3
Gemma 3 4B
- 35.3 yandexgpt-lite
- 28.3
Gemma 3 1B
- 27.8 Llama-3.2-3B-Instruct
- 25.5 Mistral-7B-Instruct-v0.3
- 19.9 Llama-3.2-1B-Instruct
- 19.5
Gemma 3 270M
- 1
Gemini 2.5 Pro Jun 2025, Thinking 84%
- 2 o3 Thinking 82.2%
- 3 GPT-5 mini Thinking 76.2%
- 4
Gemini 2.5 Flash Sep 2025, Thinking 72.8%
- 5 Claude 3.7 Sonnet Non-thinking 72.2%
- 6 GPT-4.1 Non-thinking 70.4%
- 7
Gemini 2.0 Flash Feb 2025, Non-thinking 69.6%
- 8 GPT-4o Mar 2025, Non-thinking 68.8%
- 9 GPT-5 nano Thinking 68.6%
- 10 deepseek-reasoner Thinking 68.1%
- 11 gpt-oss-120b Thinking 67.9%
- 12
Gemini 2.0 Flash-Lite Non-thinking 64.7%
- 13 GPT-4.1 mini Non-thinking 61.3%
- 14
Cohere Command A Non-thinking 59.8%
- 15 gpt-oss-20b Thinking 59.6%
- 16 deepseek-chat Non-thinking 59.1%
- 17
Llama 3.3 70B Instruct Non-thinking 57.1%
- 18 GPT-4o mini Non-thinking 57%
- 19
Llama 4 Scout Non-thinking 56.4%
- 20
Qwen3-32B Non-thinking 56.4%
- 21
Gemma 3 27B Non-thinking 56.3%
- 22 Claude 3.5 Haiku Non-thinking 54.4%
- 23
Qwen3-14B Non-thinking 51.4%
- 24
Gemma 3 12B Non-thinking 51%
- 25
Qwen3-30B-A3B Non-thinking 51%
- 26 deepseek-r1-0528-qwen3-8b Thinking 49.8%
- 27
Command R+ Non-thinking 48%
- 28 GPT-4.1 nano Non-thinking 47.8%
- 29
Mistral Small 3.1 Non-thinking 47.6%
- 30
Gemma 3n E4B Non-thinking 46.8%
- 31
Qwen3-8B Non-thinking 44.7%
- 32
Aya Expanse 32B Non-thinking 41.4%
- 33
Command R (03-2024) Non-thinking 40.5%
- 34
Qwen3-4B Non-thinking 40.3%
- 35
Gemma 3 4B Non-thinking 38.3%
- 36 yandexgpt-lite Non-thinking 35.3%
- 37
Gemma 3 1B Non-thinking 28.3%
- 38 Llama-3.2-3B-Instruct Non-thinking 27.8%
- 39 Mistral-7B-Instruct-v0.3 Non-thinking 25.5%
- 40 Llama-3.2-1B-Instruct Non-thinking 19.9%
- 41
Gemma 3 270M Non-thinking 19.5%
Source: Official PerMMLU leaderboard
- 84.0
Gemini 2.5 Pro
- 82.2 o3
- 76.2 GPT-5 mini
- 72.8
Gemini 2.5 Flash
- 72.2 Claude 3.7 Sonnet
- 70.4 GPT-4.1
- 69.6
Gemini 2.0 Flash
- 68.8 GPT-4o
- 68.6 GPT-5 nano
- 67.9 gpt-oss-120b
- 64.7
Gemini 2.0 Flash-Lite
- 61.3 GPT-4.1 mini
- 59.8
Cohere Command A
- 59.6 gpt-oss-20b
- 57.1
Llama 3.3 70B Instruct
- 57.0 GPT-4o mini
- 56.4
Llama 4 Scout
- 56.4
Qwen3-32B
- 56.3
Gemma 3 27B
- 54.4 Claude 3.5 Haiku
- 51.4
Qwen3-14B
- 51.0
Gemma 3 12B
- 51.0
Qwen3-30B-A3B
- 49.8 deepseek-r1-0528-qwen3-8b
- 48.0
Command R+
- 47.8 GPT-4.1 nano
- 47.6
Mistral Small 3.1
- 46.8
Gemma 3n E4B
- 44.7
Qwen3-8B
- 41.4
Aya Expanse 32B
- 40.5
Command R (03-2024)
- 40.3
Qwen3-4B
- 38.3
Gemma 3 4B
- 28.3
Gemma 3 1B
- 19.5
Gemma 3 270M
- 1
Gemini 2.5 Pro Jun 2025, Thinking 84%
- 2 o3 Thinking 82.2%
- 3 GPT-5 mini Thinking 76.2%
- 4
Gemini 2.5 Flash Sep 2025, Thinking 72.8%
- 5 Claude 3.7 Sonnet Non-thinking 72.2%
- 6 GPT-4.1 Non-thinking 70.4%
- 7
Gemini 2.0 Flash Feb 2025, Non-thinking 69.6%
- 8 GPT-4o Mar 2025, Non-thinking 68.8%
- 9 GPT-5 nano Thinking 68.6%
- 10 gpt-oss-120b Thinking 67.9%
- 11
Gemini 2.0 Flash-Lite Non-thinking 64.7%
- 12 GPT-4.1 mini Non-thinking 61.3%
- 13
Cohere Command A Non-thinking 59.8%
- 14 gpt-oss-20b Thinking 59.6%
- 15
Llama 3.3 70B Instruct Non-thinking 57.1%
- 16 GPT-4o mini Non-thinking 57%
- 17
Llama 4 Scout Non-thinking 56.4%
- 18
Qwen3-32B Non-thinking 56.4%
- 19
Gemma 3 27B Non-thinking 56.3%
- 20 Claude 3.5 Haiku Non-thinking 54.4%
- 21
Qwen3-14B Non-thinking 51.4%
- 22
Gemma 3 12B Non-thinking 51%
- 23
Qwen3-30B-A3B Non-thinking 51%
- 24 deepseek-r1-0528-qwen3-8b Thinking 49.8%
- 25
Command R+ Non-thinking 48%
- 26 GPT-4.1 nano Non-thinking 47.8%
- 27
Mistral Small 3.1 Non-thinking 47.6%
- 28
Gemma 3n E4B Non-thinking 46.8%
- 29
Qwen3-8B Non-thinking 44.7%
- 30
Aya Expanse 32B Non-thinking 41.4%
- 31
Command R (03-2024) Non-thinking 40.5%
- 32
Qwen3-4B Non-thinking 40.3%
- 33
Gemma 3 4B Non-thinking 38.3%
- 34
Gemma 3 1B Non-thinking 28.3%
- 35
Gemma 3 270M Non-thinking 19.5%
Source: Official PerMMLU leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 84% | Thinking | Jun 2025 | Official leaderboard |
| 2 | o3 | 82.2% | Thinking | Apr 2025 | Official leaderboard |
| 3 | GPT-5 mini | 76.2% | Thinking | Aug 2025 | Official leaderboard |
| 4 | | 72.8% | Thinking | Sep 2025 | Official leaderboard |
| 5 | Claude 3.7 Sonnet | 72.2% | Non-thinking | Feb 2025 | Official leaderboard |
| 6 | GPT-4.1 | 70.4% | Non-thinking | Apr 2025 | Official leaderboard |
| 7 | | 69.6% | Non-thinking | Feb 2025 | Official leaderboard |
| 8 | GPT-4o Mar 2025 | 68.8% | Non-thinking | Mar 2025 | Official leaderboard |
| 9 | GPT-5 nano | 68.6% | Thinking | Aug 2025 | Official leaderboard |
| 10 | deepseek-reasoner open weights, 671B | 68.1% | Thinking | Official leaderboard | |
| 11 | gpt-oss-120b | 67.9% | Thinking | Aug 2025 | Official leaderboard |
| 12 | | 64.7% | Non-thinking | Feb 2025 | Official leaderboard |
| 13 | GPT-4.1 mini | 61.3% | Non-thinking | Apr 2025 | Official leaderboard |
| 14 | | 59.8% | Non-thinking | Mar 2025 | Official leaderboard |
| 15 | gpt-oss-20b | 59.6% | Thinking | Aug 2025 | Official leaderboard |
| 16 | deepseek-chat open weights, 671B | 59.1% | Non-thinking | Official leaderboard | |
| 17 | | 57.1% | Non-thinking | Dec 2024 | Official leaderboard |
| 18 | GPT-4o mini | 57% | Non-thinking | Jul 2024 | Official leaderboard |
| 19 | | 56.4% | Non-thinking | Apr 2025 | Official leaderboard |
| 20 | | 56.4% | Non-thinking | Apr 2025 | Official leaderboard |
| 21 | | 56.3% | Non-thinking | Mar 2025 | Official leaderboard |
| 22 | Claude 3.5 Haiku | 54.4% | Non-thinking | Oct 2024 | Official leaderboard |
| 23 | | 51.4% | Non-thinking | Apr 2025 | Official leaderboard |
| 24 | | 51% | Non-thinking | Mar 2025 | Official leaderboard |
| 25 | | 51% | Non-thinking | Apr 2025 | Official leaderboard |
| 26 | deepseek-r1-0528-qwen3-8b | 49.8% | Thinking | May 2025 | Official leaderboard |
| 27 | | 48% | Non-thinking | Aug 2024 | Official leaderboard |
| 28 | GPT-4.1 nano | 47.8% | Non-thinking | Apr 2025 | Official leaderboard |
| 29 | | 47.6% | Non-thinking | Mar 2025 | Official leaderboard |
| 30 | | 46.8% | Non-thinking | Jun 2025 | Official leaderboard |
| 31 | | 44.7% | Non-thinking | Apr 2025 | Official leaderboard |
| 32 | | 41.4% | Non-thinking | Oct 2024 | Official leaderboard |
| 33 | | 40.5% | Non-thinking | Mar 2024 | Official leaderboard |
| 34 | | 40.3% | Non-thinking | Apr 2025 | Official leaderboard |
| 35 | | 38.3% | Non-thinking | Mar 2025 | Official leaderboard |
| 36 | yandexgpt-lite open weights, 8B | 35.3% | Non-thinking | Official leaderboard | |
| 37 | | 28.3% | Non-thinking | Mar 2025 | Official leaderboard |
| 38 | Llama-3.2-3B-Instruct open weights, 3.2B | 27.8% | Non-thinking | Official leaderboard | |
| 39 | Mistral-7B-Instruct-v0.3 open weights, 7.3B | 25.5% | Non-thinking | Official leaderboard | |
| 40 | Llama-3.2-1B-Instruct open weights, 1.2B | 19.9% | Non-thinking | Official leaderboard | |
| 41 | | 19.5% | Non-thinking | Aug 2025 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 84% | Thinking | Jun 2025 | Official leaderboard |
| 2 | o3 | 82.2% | Thinking | Apr 2025 | Official leaderboard |
| 3 | GPT-5 mini | 76.2% | Thinking | Aug 2025 | Official leaderboard |
| 4 | | 72.8% | Thinking | Sep 2025 | Official leaderboard |
| 5 | Claude 3.7 Sonnet | 72.2% | Non-thinking | Feb 2025 | Official leaderboard |
| 6 | GPT-4.1 | 70.4% | Non-thinking | Apr 2025 | Official leaderboard |
| 7 | | 69.6% | Non-thinking | Feb 2025 | Official leaderboard |
| 8 | GPT-4o Mar 2025 | 68.8% | Non-thinking | Mar 2025 | Official leaderboard |
| 9 | GPT-5 nano | 68.6% | Thinking | Aug 2025 | Official leaderboard |
| 10 | gpt-oss-120b | 67.9% | Thinking | Aug 2025 | Official leaderboard |
| 11 | | 64.7% | Non-thinking | Feb 2025 | Official leaderboard |
| 12 | GPT-4.1 mini | 61.3% | Non-thinking | Apr 2025 | Official leaderboard |
| 13 | | 59.8% | Non-thinking | Mar 2025 | Official leaderboard |
| 14 | gpt-oss-20b | 59.6% | Thinking | Aug 2025 | Official leaderboard |
| 15 | | 57.1% | Non-thinking | Dec 2024 | Official leaderboard |
| 16 | GPT-4o mini | 57% | Non-thinking | Jul 2024 | Official leaderboard |
| 17 | | 56.4% | Non-thinking | Apr 2025 | Official leaderboard |
| 18 | | 56.4% | Non-thinking | Apr 2025 | Official leaderboard |
| 19 | | 56.3% | Non-thinking | Mar 2025 | Official leaderboard |
| 20 | Claude 3.5 Haiku | 54.4% | Non-thinking | Oct 2024 | Official leaderboard |
| 21 | | 51.4% | Non-thinking | Apr 2025 | Official leaderboard |
| 22 | | 51% | Non-thinking | Mar 2025 | Official leaderboard |
| 23 | | 51% | Non-thinking | Apr 2025 | Official leaderboard |
| 24 | deepseek-r1-0528-qwen3-8b | 49.8% | Thinking | May 2025 | Official leaderboard |
| 25 | | 48% | Non-thinking | Aug 2024 | Official leaderboard |
| 26 | GPT-4.1 nano | 47.8% | Non-thinking | Apr 2025 | Official leaderboard |
| 27 | | 47.6% | Non-thinking | Mar 2025 | Official leaderboard |
| 28 | | 46.8% | Non-thinking | Jun 2025 | Official leaderboard |
| 29 | | 44.7% | Non-thinking | Apr 2025 | Official leaderboard |
| 30 | | 41.4% | Non-thinking | Oct 2024 | Official leaderboard |
| 31 | | 40.5% | Non-thinking | Mar 2024 | Official leaderboard |
| 32 | | 40.3% | Non-thinking | Apr 2025 | Official leaderboard |
| 33 | | 38.3% | Non-thinking | Mar 2025 | Official leaderboard |
| 34 | | 28.3% | Non-thinking | Mar 2025 | Official leaderboard |
| 35 | | 19.5% | Non-thinking | Aug 2025 | Official leaderboard |
Rankings by part
We read 18 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as cinema acc, emergency number acc and foods acc. Gemini 2.5 Pro leads 10 of them, and 4 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
cinema acc
35 models we track have a score here.
- 1
Gemini 2.5 Pro Jun 2025 93.8%
- 2
Gemini 2.5 Flash Sep 2025 ↑2 88.8%
- 3 o3 ↓1 81.3%
- 4 GPT-4.1 ↑2 76.3%
- 5 GPT-4o Mar 2025 ↑3 75%
- 6
Gemini 2.0 Flash Feb 2025 ↑1 72.5%
- 7
Gemini 2.0 Flash-Lite ↑4 71.3%
- 8 Claude 3.7 Sonnet ↓3 68.8%
- 9
Command R+ ↑16 67.5%
- 10 GPT-4o mini ↑6 66.3%
emergency number acc
35 models we track have a score here.
- 1 o3 ↑1 100%
- 2 GPT-4.1 ↑4 90%
- 3
Gemini 2.5 Pro Jun 2025 ↓2 80%
- 4
Gemini 2.5 Flash Sep 2025 80%
- 5 Claude 3.7 Sonnet 70%
- 6 GPT-4o Mar 2025 ↑2 70%
- 7 GPT-4.1 mini ↑5 70%
- 8 GPT-5 mini ↓5 70%
- 9
Gemini 2.0 Flash Feb 2025 ↓2 60%
- 10
Gemini 2.0 Flash-Lite ↑1 60%
foods acc
35 models we track have a score here.
- 1
Gemini 2.5 Flash Sep 2025 ↑3 82%
- 2
Gemini 2.5 Pro Jun 2025 ↓1 80%
- 3
Gemini 2.0 Flash Feb 2025 ↑4 79%
- 4 GPT-4o Mar 2025 ↑4 78%
- 5 GPT-4.1 ↑1 78%
- 6 o3 ↓4 77%
- 7 GPT-5 mini ↓4 77%
- 8 Claude 3.7 Sonnet ↓3 74%
- 9 GPT-4.1 mini ↑3 73%
- 10 Claude 3.5 Haiku ↑10 72%
games acc
35 models we track have a score here.
- 1 o3 ↑1 75%
- 2
Gemma 3 4B ↑31 70%
- 3 GPT-4.1 ↑3 70%
- 4
Cohere Command A ↑9 65%
- 5
Gemma 3n E4B ↑23 65%
- 6
Llama 3.3 70B Instruct ↑9 60%
- 7 GPT-4o Mar 2025 ↑1 60%
- 8 GPT-4.1 nano ↑18 60%
- 9 GPT-5 mini ↓6 60%
- 10 Claude 3.5 Haiku ↑10 55%
herbal drugs acc
35 models we track have a score here.
- 1
Gemini 2.0 Flash Feb 2025 ↑6 75%
- 2
Gemini 2.5 Pro Jun 2025 ↓1 75%
- 3
Gemini 2.5 Flash Sep 2025 ↑1 75%
- 4
Gemini 2.0 Flash-Lite ↑7 70%
- 5 GPT-4o Mar 2025 ↑3 70%
- 6 GPT-5 mini ↓3 70%
- 7 GPT-5 nano ↑2 70%
- 8 GPT-4o mini ↑8 67.5%
- 9
Aya Expanse 32B ↑21 67.5%
- 10 Claude 3.7 Sonnet ↓5 67.5%
places acc
35 models we track have a score here.
- 1 o3 ↑1 89.5%
- 2
Gemini 2.5 Flash Sep 2025 ↑2 89.5%
- 3
Gemini 2.5 Pro Jun 2025 ↓2 88.6%
- 4 GPT-5 mini ↓1 88.6%
- 5
Gemini 2.0 Flash-Lite ↑6 86.7%
- 6 GPT-4.1 86.7%
- 7 Claude 3.7 Sonnet ↓2 84.8%
- 8
Cohere Command A ↑5 84.8%
- 9
Gemini 2.0 Flash Feb 2025 ↓2 83.8%
- 10 GPT-4o Mar 2025 ↓2 83.8%
poetry acc
35 models we track have a score here.
- 1 GPT-4o Mar 2025 ↑7 90%
- 2
Gemini 2.5 Pro Jun 2025 ↓1 90%
- 3
Gemini 2.5 Flash Sep 2025 ↑1 90%
- 4 GPT-4.1 ↑2 87.5%
- 5 o3 ↓3 87.5%
- 6
Gemini 2.0 Flash Feb 2025 ↑1 82.5%
- 7
Gemini 2.0 Flash-Lite ↑4 80%
- 8 Claude 3.7 Sonnet ↓3 80%
- 9
Command R+ ↑16 77.5%
- 10
Cohere Command A ↑3 77.5%
politicians acc
35 models we track have a score here.
- 1 o3 ↑1 100%
- 2 Claude 3.7 Sonnet ↑3 95%
- 3
Cohere Command A ↑10 95%
- 4 GPT-4o Mar 2025 ↑4 95%
- 5
Gemini 2.5 Pro Jun 2025 ↓4 95%
- 6
Gemini 2.5 Flash Sep 2025 ↓2 90%
- 7
Llama 3.3 70B Instruct ↑8 85%
- 8
Llama 4 Scout ↑9 85%
- 9 GPT-4.1 ↓3 85%
- 10
Gemini 2.0 Flash-Lite ↑1 80%
popular people acc
35 models we track have a score here.
- 1
Gemini 2.5 Pro Jun 2025 92.8%
- 2 o3 91.8%
- 3 Claude 3.7 Sonnet ↑2 86.2%
- 4 GPT-4o Mar 2025 ↑4 86.2%
- 5
Gemini 2.5 Flash Sep 2025 ↓1 85.6%
- 6 GPT-4.1 84.6%
- 7 GPT-5 mini ↓4 81.5%
- 8
Cohere Command A ↑5 80.9%
- 9
Gemini 2.0 Flash Feb 2025 ↓2 78.5%
- 10
Gemini 2.0 Flash-Lite ↑1 77.4%
Government law acc
35 models we track have a score here.
- 1
Gemini 2.5 Pro Jun 2025 97.8%
- 2
Gemini 2.0 Flash Feb 2025 ↑5 95.7%
- 3
Gemini 2.5 Flash Sep 2025 ↑1 95.7%
- 4
Gemini 2.0 Flash-Lite ↑7 93.5%
- 5 GPT-4o Mar 2025 ↑3 93.5%
- 6 GPT-5 mini ↓3 93.5%
- 7 GPT-5 nano ↑2 93.5%
- 8 GPT-4o mini ↑8 89.1%
- 9
Llama 3.3 70B Instruct ↑6 89.1%
- 10 Claude 3.7 Sonnet ↓5 89.1%
proverbs acc
35 models we track have a score here.
- 1
Gemini 2.5 Pro Jun 2025 90%
- 2 o3 89%
- 3
Gemini 2.5 Flash Sep 2025 ↑1 87%
- 4 GPT-4.1 ↑2 86%
- 5 GPT-4o Mar 2025 ↑3 80%
- 6
Gemini 2.0 Flash Feb 2025 ↑1 78%
- 7
Cohere Command A ↑6 78%
- 8
Gemini 2.0 Flash-Lite ↑3 77%
- 9 Claude 3.7 Sonnet ↓4 77%
- 10 GPT-4o mini ↑6 75%
religous acc
35 models we track have a score here.
- 1
Gemini 2.5 Pro Jun 2025 97.8%
- 2 o3 95.6%
- 3
Gemini 2.5 Flash Sep 2025 ↑1 95.6%
- 4 Claude 3.7 Sonnet ↑1 93.3%
- 5 GPT-4o Mar 2025 ↑3 93.3%
- 6
Gemini 2.0 Flash Feb 2025 ↑1 91.1%
- 7
Gemini 2.0 Flash-Lite ↑4 91.1%
- 8 GPT-4.1 ↓2 88.9%
- 9 GPT-5 mini ↓6 88.9%
- 10
Cohere Command A ↑3 86.7%
social manners acc
35 models we track have a score here.
- 1
Gemini 2.5 Pro Jun 2025 94.4%
- 2 Claude 3.7 Sonnet ↑3 91%
- 3
Gemini 2.5 Flash Sep 2025 ↑1 91%
- 4
Cohere Command A ↑9 89.9%
- 5 o3 ↓3 88.8%
- 6
Gemma 3 27B ↑13 87.6%
- 7 GPT-4.1 ↓1 86.5%
- 8 Claude 3.5 Haiku ↑12 85.4%
- 9
Llama 3.3 70B Instruct ↑6 85.4%
- 10
Gemini 2.0 Flash Feb 2025 ↓3 85.4%
souvenirs acc
35 models we track have a score here.
- 1 o3 ↑1 78%
- 2
Gemini 2.5 Pro Jun 2025 ↓1 78%
- 3
Gemini 2.0 Flash Feb 2025 ↑4 74%
- 4 GPT-5 mini ↓1 74%
- 5 Claude 3.7 Sonnet 72%
- 6
Llama 4 Scout ↑11 72%
- 7
Gemini 2.5 Flash Sep 2025 ↓3 72%
- 8 GPT-4o mini ↑8 68%
- 9
Gemini 2.0 Flash-Lite ↑2 68%
- 10
Cohere Command A ↑3 68%
sports acc
35 models we track have a score here.
- 1
Gemini 2.5 Pro Jun 2025 65.1%
- 2 Claude 3.7 Sonnet ↑3 60.3%
- 3
Gemma 3 27B ↑16 60.3%
- 4
Gemini 2.0 Flash-Lite ↑7 58.7%
- 5 GPT-5 nano ↑4 58.7%
- 6
Gemini 2.5 Flash Sep 2025 ↓2 58.7%
- 7
Gemma 3 12B ↑15 57.1%
- 8 o3 ↓6 57.1%
- 9 gpt-oss-120b ↑1 57.1%
- 10 GPT-4o mini ↑6 55.6%
GPK acc
35 models we track have a score here.
- 1
Gemini 2.5 Pro Jun 2025 87.3%
- 2 o3 84.5%
- 3
Gemini 2.5 Flash Sep 2025 ↑1 83.9%
- 4 GPT-4o Mar 2025 ↑4 80.2%
- 5 GPT-4.1 ↑1 80.1%
- 6 Claude 3.7 Sonnet ↓1 79.4%
- 7
Gemini 2.0 Flash Feb 2025 77.6%
- 8
Gemini 2.0 Flash-Lite ↑3 76.9%
- 9 GPT-5 mini ↓6 76.7%
- 10
Cohere Command A ↑3 76%
SPK acc
35 models we track have a score here.
- 1
Gemini 2.5 Pro Jun 2025 85.6%
- 2 o3 83.4%
- 3 GPT-5 mini 76.7%
- 4
Gemini 2.5 Flash Sep 2025 76.5%
- 5 Claude 3.7 Sonnet 75.9%
- 6
Gemini 2.0 Flash Feb 2025 ↑1 72.6%
- 7 GPT-4.1 ↓1 72.6%
- 8 GPT-4o Mar 2025 72%
- 9 GPT-5 nano 68.7%
- 10 gpt-oss-120b 67.7%
UPK acc
35 models we track have a score here.
- 1
Gemini 2.5 Pro Jun 2025 82.4%
- 2 o3 81%
- 3 GPT-5 mini 75.8%
- 4
Gemini 2.5 Flash Sep 2025 68.6%
- 5 Claude 3.7 Sonnet 68.5%
- 6 GPT-5 nano ↑3 68.5%
- 7 gpt-oss-120b ↑3 68.5%
- 8 GPT-4.1 ↓2 67.6%
- 9
Gemini 2.0 Flash Feb 2025 ↓2 66.4%
- 10 GPT-4o Mar 2025 ↓2 65.1%
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.
Source: Official PerMMLU leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| Thinking | 84% | ||||
| o3 | Thinking | 82.2% | |||
| GPT-5 mini | Thinking | 76.2% | |||
| Thinking | 72.8% | ||||
| Claude 3.7 Sonnet | Non-thinking | 72.2% | |||
| GPT-4.1 | Non-thinking | 70.4% | |||
| Non-thinking | 69.6% | ||||
| GPT-4o Mar 2025 | Non-thinking | 68.8% | |||
| GPT-5 nano | Thinking | 68.6% | |||
| gpt-oss-120b | Thinking | 67.9% | |||
| Non-thinking | 64.7% | ||||
| gpt-oss-120b | low reasoning | 62.2% | |||
| GPT-4.1 mini | Non-thinking | 61.3% | |||
| Non-thinking | 59.8% | ||||
| gpt-oss-20b | Thinking | 59.6% | |||
| GPT-5 mini | minimal reasoning | 59.5% | |||
| Non-thinking | 57.1% | ||||
| GPT-4o mini | Non-thinking | 57% | |||
| Non-thinking | 56.4% | ||||
| Non-thinking | 56.4% | ||||
| Non-thinking | 56.3% | ||||
| Claude 3.5 Haiku | Non-thinking | 54.4% | |||
| gpt-oss-20b | low reasoning | 54.4% | |||
| Non-thinking | 51.4% | ||||
| Non-thinking | 51% | ||||
| Non-thinking | 51% | ||||
| deepseek-r1-0528-qwen3-8b | Thinking | 49.8% | |||
| Non-thinking | 48% | ||||
| GPT-4.1 nano | Non-thinking | 47.8% | |||
| Non-thinking | 47.6% | ||||
| Non-thinking | 46.8% | ||||
| GPT-5 nano | minimal reasoning | 45.6% | |||
| Non-thinking | 44.7% | ||||
| Non-thinking | 41.4% | ||||
| Non-thinking | 40.5% | ||||
| Non-thinking | 40.3% | ||||
| Non-thinking | 38.3% | ||||
| Non-thinking | 28.3% | ||||
| Non-thinking | 19.5% |
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.
Source: Official PerMMLU leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| Thinking | 84% | ||||
| o3 | Thinking | 82.2% | |||
| GPT-5 mini | Thinking | 76.2% | |||
| Thinking | 72.8% | ||||
| Claude 3.7 Sonnet | Non-thinking | 72.2% | |||
| GPT-4.1 | Non-thinking | 70.4% | |||
| Non-thinking | 69.6% | ||||
| GPT-4o Mar 2025 | Non-thinking | 68.8% | |||
| GPT-5 nano | Thinking | 68.6% | |||
| gpt-oss-120b | Thinking | 67.9% | |||
| Non-thinking | 64.7% | ||||
| gpt-oss-120b | low reasoning | 62.2% | |||
| GPT-4.1 mini | Non-thinking | 61.3% | |||
| Non-thinking | 59.8% | ||||
| gpt-oss-20b | Thinking | 59.6% | |||
| GPT-5 mini | minimal reasoning | 59.5% | |||
| Non-thinking | 57.1% | ||||
| GPT-4o mini | Non-thinking | 57% | |||
| Non-thinking | 56.4% | ||||
| Non-thinking | 56.4% | ||||
| Non-thinking | 56.3% | ||||
| Claude 3.5 Haiku | Non-thinking | 54.4% | |||
| gpt-oss-20b | low reasoning | 54.4% | |||
| Non-thinking | 51.4% | ||||
| Non-thinking | 51% | ||||
| Non-thinking | 51% | ||||
| deepseek-r1-0528-qwen3-8b | Thinking | 49.8% | |||
| Non-thinking | 48% | ||||
| GPT-4.1 nano | Non-thinking | 47.8% | |||
| Non-thinking | 47.6% | ||||
| Non-thinking | 46.8% | ||||
| GPT-5 nano | minimal reasoning | 45.6% | |||
| Non-thinking | 44.7% | ||||
| Non-thinking | 41.4% | ||||
| Non-thinking | 40.5% | ||||
| Non-thinking | 40.3% | ||||
| Non-thinking | 38.3% | ||||
| Non-thinking | 28.3% | ||||
| Non-thinking | 19.5% |
2 models have held number one on the official leaderboard since its release in May 2025
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official PerMMLU leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official PerMMLU 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
- 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
9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 14 months ago, in September 2025, for Gemini 2.5 Flash. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
- Pass Independent runs Scores from the official leaderboard (41 entries).
- Pass Models scored 35 models we track, plus 6 other entries on its leaderboard.
- Warning Newest score The newest came 14 months ago, in September 2025, for Gemini 2.5 Flash.
- Note Frontier models None of the top 10 has a score. The strongest tested is o3 (#84).
- Pass Headroom The best is 84%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on mcinext-mizan-llm-leaderboard.hf.space: 39 rows for models we track and 6 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.