IndoMMLU
LLaMA-65B 35.8% leads the models we track 
Qwen2.5-72B 69.3% leads since September 2024 Models answer 14,981 multiple-choice questions from Indonesian school and university entrance exams, including nine local languages. Scored by accuracy.
- Made by
- MBZUAI
- Released
- October 2023
- Kind
- Knowledge in many languages
- Level
- School
- Scored
- Checked automatically
- Answers
- Multiple choice
- Tasks
- All public
- Strongest model tested
- Qwen2.5-72B #165 on the capability index
- Last in a launch post
- September 2024, dropped before it saturated
- Last model added
- December 2024
- Scores refreshed
- November 12th, 2026
In the authors’ words
Although large language models (LLMs) are often pre-trained on large-scale multilingual texts, their reasoning abilities and real-world knowledge are mainly evaluated based on English datasets. Assessing LLM capabilities beyond English is increasingly vital but hindered due to the lack of suitable datasets. In this work, we introduce IndoMMLU, the first multi-task language understanding benchmark for Indonesian culture and languages, which consists of questions from primary school to university entrance exams in Indonesia. By employing professional teachers, we obtain 14,981 questions across 64 tasks and education levels, with 46% of the questions focusing on assessing proficiency in the Indonesian language and knowledge of nine local languages and cultures in Indonesia. Our empirical evaluations show that GPT-3.5 only manages to pass the Indonesian primary school level, with limited knowledge of local Indonesian languages and culture. Other smaller models such as BLOOMZ and Falcon perform at even lower levels.
GPT-3.5 (175B) tops the leaderboard with 53.2%
Best score per entry from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best. 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 LLaMA-65B, at 35.8%.
Qwen2.5-72B leads with 69.3%, claims included
Best score per model from every source, including labs’ own numbers and claims Epoch AI collected from reports and papers. Labs pick their own settings, so a claim can beat an independent run of the same model.
The newest model with a score here came out in April 2023, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- Claude Fable 5.2
-
Gemini 4 Carbon - GPT-6 Astra
- Claude Opus 5.5
- and 16 more
- 53.2 GPT-3.5 (175B)
- 42.5 mT0 xxl (13B)
- 38.2 mT0 xl (3.7B)
- 38.0 BLOOMZ (7.1B)
- 36.4 BLOOMZ (3B)
- 35.8 LLaMA-65B
- 34.4 BLOOMZ (1.7B)
- 32.1
Falcon-40B - 30.3 Bactrian-X-LLaMA (13B)
- 26.5 LLaMA (30B)
- 25.7 Bactrian-X-LLaMA (7B)
- 25.3 LLaMA-13B
- 25.3 LLaMA-7B
- 25.2 mT0 large (1.2B)
- 25.2 XGLM (2.9B)
- 25.2 XGLM (564M)
- 25.1 Falcon-7B
- 25.0 mT0 base (580M)
- 25.0 XGLM (4.5B)
- 24.9 mT0 small (300M)
- 24.5 XGLM (7.5B)
- 24.4 XGLM (1.7B)
- 24.0 BLOOMZ (560M)
- 22.4 BLOOMZ (1.1B)
- 1 GPT-3.5 (175B) zero-shot 53.2%
- 2 mT0 xxl (13B) zero-shot 42.5%
- 3 mT0 xl (3.7B) zero-shot 38.2%
- 4 BLOOMZ (7.1B) zero-shot 38%
- 5 BLOOMZ (3B) zero-shot 36.4%
- 6 LLaMA-65B zero-shot 35.8%
- 7 BLOOMZ (1.7B) zero-shot 34.4%
- 8 Falcon-40B zero-shot 32.1%
- 9 Bactrian-X-LLaMA (13B) zero-shot 30.3%
- 10 LLaMA (30B) zero-shot 26.5%
- 11 Bactrian-X-LLaMA (7B) zero-shot 25.7%
- 12 LLaMA-13B zero-shot 25.3%
- 13 LLaMA-7B zero-shot 25.3%
- 14 mT0 large (1.2B) zero-shot 25.2%
- 15 XGLM (2.9B) zero-shot 25.2%
- 16 XGLM (564M) zero-shot 25.2%
- 17 Falcon-7B zero-shot 25.1%
- 18 mT0 base (580M) zero-shot 25%
- 19 XGLM (4.5B) zero-shot 25%
- 20 mT0 small (300M) zero-shot 24.9%
- 21 XGLM (7.5B) zero-shot 24.5%
- 22 XGLM (1.7B) zero-shot 24.4%
- 23 BLOOMZ (560M) zero-shot 24%
- 24 BLOOMZ (1.1B) zero-shot 22.4%
Source: Official IndoMMLU leaderboard
- 1 Qwen2.5-72B 69.3%
- 2 Qwen2.5-32B 66.4%
- 3 Qwen2.5-14B 65.1%
- 4 Qwen2.5-7B 56.4%
- 5 LLaMA-65B zero-shot 35.8%
- 6 Falcon-40B zero-shot 32.1%
- 7 LLaMA-13B zero-shot 25.3%
- 8 LLaMA-7B zero-shot 25.3%
- 9 Falcon-7B zero-shot 25.1%
Sources: Labs’ model cards and official IndoMMLU leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GPT-3.5 (175B) | 53.2% | zero-shot | Official leaderboard | |
| 2 | mT0 xxl (13B) | 42.5% | zero-shot | Official leaderboard | |
| 3 | mT0 xl (3.7B) | 38.2% | zero-shot | Official leaderboard | |
| 4 | BLOOMZ (7.1B) | 38% | zero-shot | Official leaderboard | |
| 5 | BLOOMZ (3B) | 36.4% | zero-shot | Official leaderboard | |
| 6 | LLaMA-65B | 35.8% | zero-shot | Feb 2023 | Official leaderboard |
| 7 | BLOOMZ (1.7B) | 34.4% | zero-shot | Official leaderboard | |
| 8 | Falcon-40B | 32.1% | zero-shot | Mar 2023 | Official leaderboard |
| 9 | Bactrian-X-LLaMA (13B) | 30.3% | zero-shot | Official leaderboard | |
| 10 | LLaMA (30B) | 26.5% | zero-shot | Official leaderboard | |
| 11 | Bactrian-X-LLaMA (7B) | 25.7% | zero-shot | Official leaderboard | |
| 12 | LLaMA-13B | 25.3% | zero-shot | Feb 2023 | Official leaderboard |
| 13 | LLaMA-7B | 25.3% | zero-shot | Feb 2023 | Official leaderboard |
| 14 | mT0 large (1.2B) | 25.2% | zero-shot | Official leaderboard | |
| 15 | XGLM (2.9B) | 25.2% | zero-shot | Official leaderboard | |
| 16 | XGLM (564M) | 25.2% | zero-shot | Official leaderboard | |
| 17 | Falcon-7B | 25.1% | zero-shot | Apr 2023 | Official leaderboard |
| 18 | mT0 base (580M) | 25% | zero-shot | Official leaderboard | |
| 19 | XGLM (4.5B) | 25% | zero-shot | Official leaderboard | |
| 20 | mT0 small (300M) | 24.9% | zero-shot | Official leaderboard | |
| 21 | XGLM (7.5B) | 24.5% | zero-shot | Official leaderboard | |
| 22 | XGLM (1.7B) | 24.4% | zero-shot | Official leaderboard | |
| 23 | BLOOMZ (560M) | 24% | zero-shot | Official leaderboard | |
| 24 | BLOOMZ (1.1B) | 22.4% | zero-shot | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Qwen2.5-72B | 69.3% | Sep 2024 | Lab model card | |
| 2 | Qwen2.5-32B | 66.4% | Dec 2024 | Lab model card | |
| 3 | Qwen2.5-14B | 65.1% | Dec 2024 | Lab model card | |
| 4 | Qwen2.5-7B | 56.4% | Dec 2024 | Lab model card | |
| 5 | LLaMA-65B | 35.8% | zero-shot | Feb 2023 | Official leaderboard |
| 6 | Falcon-40B | 32.1% | zero-shot | Mar 2023 | Official leaderboard |
| 7 | LLaMA-13B | 25.3% | zero-shot | Feb 2023 | Official leaderboard |
| 8 | LLaMA-7B | 25.3% | zero-shot | Feb 2023 | Official leaderboard |
| 9 | Falcon-7B | 25.1% | zero-shot | Apr 2023 | Official leaderboard |
Labs report scores for 4 models in their cards. Switch the scores above to rank them together.
Number one over time 2 models have held number one with claims included since its release in October 2023
Lab claims count from the day the card came out, independent runs from the model’s release. We draw one row each for the benchmark’s own leaderboard and what labs claim in their cards.
Sources: Official IndoMMLU leaderboard and labs’ model cards
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Sources: Official IndoMMLU leaderboard and labs’ model cards
No launch post has listed it since September 2024
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.
| Date | Launch post | Score | As the post writes it |
|---|---|---|---|
| Sep 19, 2024 | Qwen2.5-72B | 69.3 | IndoMMLU (Indonesian) |
4 system cards report it, from 1 lab
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.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| Dec 19, 2024 | Qwen2.5-7B technical report | 56.4 | IndoMMLU (Indonesian), Qwen2.5-7B-Instruct (Table 14) | Capabilities |
| Dec 19, 2024 | Qwen2.5-72B technical report | 69.3 | IndoMMLU (Indonesian), Qwen2.5-72B-Instruct (Table 13) | Capabilities |
| Dec 19, 2024 | Qwen2.5-32B technical report | 66.4 | IndoMMLU (Indonesian), Qwen2.5-32B-Instruct (Table 13) | Capabilities |
| Dec 19, 2024 | Qwen2.5-14B technical report | 65.1 | IndoMMLU (Indonesian), Qwen2.5-14B-Instruct (Table 14) | Capabilities |
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.
Nemotron 3 Ultra 550B-A55B Base 90.1% - 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%