Benchmark Almanac

Open Persian LLM Leaderboard

Llama 3.3 70B Instruct 70.4% leads since December 2024
Not in model cards

Models answer Persian multiple-choice questions. The leaderboard averages accuracy across five tasks, including two question sets from the organizers.

The project description also lists GSM Persian. The scoring code omits it, and the published results contain five task scores for each model.

Released
November 2024
Kind
Knowledge in many languages
Level
University
Scored
Checked automatically
Answers
Multiple choice
Tasks
Public, plus a private test set
Measures
Capability
Language
Persian
In and out
Text → Text
Interaction
Single turn
Refresh
One fixed set
Run by
A host
Strongest model tested
Gemma 3 27B #165 on the capability index
Last model added
March 2025
Scores refreshed
November 20th, 2026
29
Entries on its own leaderboard, 17 models we track

Llama 3.3 70B Instruct leads with 70.4% among models released by March 2025 Llama 3.3 70B Instruct leads with 70.4% among models released by March 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. 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 March 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 Llama 3.3 70B Instruct bfloat16 70.4%
  2. 2 Gemma 3 27B bfloat16 67.8%
  3. 3 Gemma 2 27B bfloat16 65.5%
  4. 4 QwQ-32B-Preview bfloat16 64.8%
  5. 5 Qwen2.5-32B bfloat16 64.5%
  6. 6 Gemma 2 9B bfloat16 62.9%
  7. 7 Aya Expanse 32B float16 61.9%
  8. 8 Gemma 3 12B bfloat16 60.9%
  9. 9 CohereForAI/aya-23-35B float16 56.8%
  10. 10 Aya Expanse 8B float16 53.7%
  11. 11 Qwen2.5-7B bfloat16 51.9%
  12. 12 Qwen2-7B-Instruct bfloat16 51.4%
  13. 13 PartAI/Dorna2-Llama3.1-8B-Instruct bfloat16 50.7%
  14. 14 O1-OPEN/OpenO1-LLama-8B-v0.1 bfloat16 50.2%
  15. 15 Llama 3.1-8B bfloat16 50.1%
  16. 16 CohereForAI/aya-23-8B float16 49.8%
  17. 17 PrincetonPLI/Llama-3-8B-Instruct-SkillMix bfloat16 49.1%
  18. 18 Gemma 3 4B bfloat16 49.1%
  19. 19 Llama 3-8B bfloat16 49%
  20. 20 NousResearch/Hermes-3-Llama-3.1-8B bfloat16 48.8%
  21. 21 PartAI/Dorna-Llama3-8B-Instruct bfloat16 48.3%
  22. 22 lightblue/suzume-llama-3-8B-multilingual bfloat16 47.3%
  23. 23 Gemma 2 2B IT bfloat16 45.3%
  24. 24 amirakhlaghiqqq/PersianPhi bfloat16 41.5%
  25. 25 universitytehran/PersianMind-v1.0 bfloat16 39.5%
  26. 26 Skywork/Skywork-o1-Open-Llama-3.1-8B bfloat16 34.1%
  27. 27 Llama 3.2 1B bfloat16 32.7%
  28. 28 MaralGPT/Maral-7B-alpha-1 bfloat16 30.8%
  29. 29 Gemma 3 1B bfloat16 30.7%

Source: Official Open Persian LLM Leaderboard leaderboard

#EntryScoreSettingDateLinkSource
1 Llama 3.3 70B Instruct 70.4% bfloat16 Dec 2024 huggingface.co Official leaderboard
2 Gemma 3 27B 67.8% bfloat16 Mar 2025 huggingface.co Official leaderboard
3 Gemma 2 27B 65.5% bfloat16 Jun 2024 huggingface.co Official leaderboard
4 QwQ-32B-Preview 64.8% bfloat16 Nov 2024 huggingface.co Official leaderboard
5 Qwen2.5-32B 64.5% bfloat16 Sep 2024 huggingface.co Official leaderboard
6 Gemma 2 9B 62.9% bfloat16 Jun 2024 huggingface.co Official leaderboard
7 Aya Expanse 32B 61.9% float16 Oct 2024 huggingface.co Official leaderboard
8 Gemma 3 12B 60.9% bfloat16 Mar 2025 huggingface.co Official leaderboard
9 CohereForAI/aya-23-35B 35B 56.8% float16 huggingface.co Official leaderboard
10 Aya Expanse 8B 53.7% float16 Oct 2024 huggingface.co Official leaderboard
11 Qwen2.5-7B 51.9% bfloat16 Sep 2024 huggingface.co Official leaderboard
12 Qwen2-7B-Instruct 51.4% bfloat16 Jun 2024 huggingface.co Official leaderboard
13 PartAI/Dorna2-Llama3.1-8B-Instruct 8B 50.7% bfloat16 huggingface.co Official leaderboard
14 O1-OPEN/OpenO1-LLama-8B-v0.1 8B 50.2% bfloat16 huggingface.co Official leaderboard
15 Llama 3.1-8B 50.1% bfloat16 Jul 2024 huggingface.co Official leaderboard
16 CohereForAI/aya-23-8B 8B 49.8% float16 huggingface.co Official leaderboard
17 PrincetonPLI/Llama-3-8B-Instruct-SkillMix 8B 49.1% bfloat16 huggingface.co Official leaderboard
18 Gemma 3 4B 49.1% bfloat16 Mar 2025 huggingface.co Official leaderboard
19 Llama 3-8B 49% bfloat16 Apr 2024 huggingface.co Official leaderboard
20 NousResearch/Hermes-3-Llama-3.1-8B 8B 48.8% bfloat16 huggingface.co Official leaderboard
21 PartAI/Dorna-Llama3-8B-Instruct 8B 48.3% bfloat16 huggingface.co Official leaderboard
22 lightblue/suzume-llama-3-8B-multilingual 8B 47.3% bfloat16 huggingface.co Official leaderboard
23 Gemma 2 2B IT 45.3% bfloat16 Jul 2024 huggingface.co Official leaderboard
24 amirakhlaghiqqq/PersianPhi 3.9B 41.5% bfloat16 huggingface.co Official leaderboard
25 universitytehran/PersianMind-v1.0 6.8B 39.5% bfloat16 huggingface.co Official leaderboard
26 Skywork/Skywork-o1-Open-Llama-3.1-8B 8B 34.1% bfloat16 huggingface.co Official leaderboard
27 Llama 3.2 1B 32.7% bfloat16 Sep 2024 huggingface.co Official leaderboard
28 MaralGPT/Maral-7B-alpha-1 7.2B 30.8% bfloat16 huggingface.co Official leaderboard
29 Gemma 3 1B 30.7% bfloat16 Mar 2025 huggingface.co Official leaderboard
#ModelScoreSettingDateLinkSource
1 Llama 3.3 70B Instruct 70.4% bfloat16 Dec 2024 huggingface.co Official leaderboard
2 Gemma 3 27B 67.8% bfloat16 Mar 2025 huggingface.co Official leaderboard
3 Gemma 2 27B 65.5% bfloat16 Jun 2024 huggingface.co Official leaderboard
4 QwQ-32B-Preview 64.8% bfloat16 Nov 2024 huggingface.co Official leaderboard
5 Qwen2.5-32B 64.5% bfloat16 Sep 2024 huggingface.co Official leaderboard
6 Gemma 2 9B 62.9% bfloat16 Jun 2024 huggingface.co Official leaderboard
7 Aya Expanse 32B 61.9% float16 Oct 2024 huggingface.co Official leaderboard
8 Gemma 3 12B 60.9% bfloat16 Mar 2025 huggingface.co Official leaderboard
9 Aya Expanse 8B 53.7% float16 Oct 2024 huggingface.co Official leaderboard
10 Qwen2.5-7B 51.9% bfloat16 Sep 2024 huggingface.co Official leaderboard
11 Qwen2-7B-Instruct 51.4% bfloat16 Jun 2024 huggingface.co Official leaderboard
12 Llama 3.1-8B 50.1% bfloat16 Jul 2024 huggingface.co Official leaderboard
13 Gemma 3 4B 49.1% bfloat16 Mar 2025 huggingface.co Official leaderboard
14 Llama 3-8B 49% bfloat16 Apr 2024 huggingface.co Official leaderboard
15 Gemma 2 2B IT 45.3% bfloat16 Jul 2024 huggingface.co Official leaderboard
16 Llama 3.2 1B 32.7% bfloat16 Sep 2024 huggingface.co Official leaderboard
17 Gemma 3 1B 30.7% bfloat16 Mar 2025 huggingface.co Official leaderboard

Rankings by part

We read 5 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Part Multiple Choice, ARC Easy and ARC Challenge. Llama 3.3 70B Instruct leads 3 of them, and 2 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.

Part Multiple Choice

17 models we track have a score here.

  1. 1 Llama 3.3 70B Instruct 51.4%
  2. 2 Gemma 3 27B 48.6%
  3. 3 QwQ-32B-Preview ↑1 46.6%
  4. 4 Qwen2.5-32B ↑1 46.1%
  5. 5 Gemma 2 27B ↓2 46%
  6. 6 Gemma 3 12B ↑2 43.5%
  7. 7 Aya Expanse 32B 43.4%
  8. 8 Gemma 2 9B ↓2 42.7%
  9. 9 Qwen2.5-7B ↑1 36.7%
  10. 10 Llama 3.1-8B ↑2 36.7%

2 models have held number one on the official leaderboard since its release in November 2024

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

Google DeepMind Meta
Own leaderboard Not tested since Mar 2025

Source: Official Open Persian LLM Leaderboard leaderboard

Scores over time

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 Near the ceiling (85%) Llama 3.3 70B Instruct Gemma 2 27B Qwen2-7B-Instruct 0% 20% 40% 60% 80% 100% 2024 2025 Near the ceiling (85%) Llama 3.3 70B Instruct Qwen2-7B-Instruct

Source: Official Open Persian LLM Leaderboard 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

9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 20 months ago, in March 2025, for Gemma 3 12B. Every benchmark

Catalog record 2 pass
  • Pass Task access Public tasks, plus a private test set.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
  • Pass Independent runs Scores from the official leaderboard (29 entries).
  • Pass Models scored 17 models we track, plus 12 other entries on its leaderboard.
  • Warning Newest score The newest came 20 months ago, in March 2025, for Gemma 3 12B.
  • Note Frontier models None of the top 10 has a score. The strongest tested is Gemma 3 27B (#165).
  • Pass Headroom The best is 70.4%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on huggingface.co: 19 rows for models we track and 12 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