Benchmark Almanac

PerCoR

Gemini 2.5 Pro Jun 2025 94.3% leads since June 2025
Not in model cards

Models read the beginning of a Persian passage and choose its original ending from four options.

MIZAN reports accuracy for 45 model entries. The original test release contains 10,000 questions.

Made by
MCINext
Released
May 2025
Kind
Reasoning in many languages
Level
Everyday
Scored
Checked automatically
Answers
Multiple choice
Tasks
All public
Measures
Capability
Language
Persian
In and out
Text → Text
Interaction
Single turn
Items from
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
41
Entries on its own leaderboard, 35 models we track

Gemini 2.5 Pro leads with 94.3% among models released by September 2025 Gemini 2.5 Pro leads with 94.3% 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:

  1. 1 Gemini 2.5 Pro Jun 2025, Thinking 94.3%
  2. 2 o3 Thinking 92.2%
  3. 3 Claude 3.7 Sonnet Non-thinking 91.2%
  4. 4 GPT-4.1 Non-thinking 88.4%
  5. 5 Gemini 2.5 Flash Sep 2025, Thinking 87.2%
  6. 6 GPT-4o Mar 2025, Non-thinking 86.7%
  7. 7 GPT-5 mini Thinking 86.5%
  8. 8 Gemini 2.0 Flash Feb 2025, Non-thinking 86.4%
  9. 9 deepseek-reasoner Thinking 82.5%
  10. 10 deepseek-chat Non-thinking 82.4%
  11. 11 Gemini 2.0 Flash-Lite Non-thinking 81.4%
  12. 12 Cohere Command A Non-thinking 79.9%
  13. 13 Llama 3.3 70B Instruct Non-thinking 79.6%
  14. 14 Llama 4 Scout Non-thinking 78.1%
  15. 15 gpt-oss-120b Thinking 77.1%
  16. 16 GPT-4.1 mini Non-thinking 77.1%
  17. 17 Qwen3-32B Non-thinking 76.5%
  18. 18 Gemma 3 27B Non-thinking 76.3%
  19. 19 GPT-4o mini Non-thinking 76%
  20. 20 Command R+ Non-thinking 73.6%
  21. 21 GPT-5 nano Thinking 73.1%
  22. 22 Claude 3.5 Haiku Non-thinking 71.6%
  23. 23 Gemma 3 12B Non-thinking 70.9%
  24. 24 gpt-oss-20b Thinking 69.7%
  25. 25 Qwen3-14B Non-thinking 69.6%
  26. 26 Mistral Small 3.1 Non-thinking 68.9%
  27. 27 Qwen3-30B-A3B Non-thinking 68.8%
  28. 28 Aya Expanse 32B Non-thinking 63.3%
  29. 29 Command R (03-2024) Non-thinking 60%
  30. 30 Gemma 3n E4B Non-thinking 59.1%
  31. 31 GPT-4.1 nano Non-thinking 54.9%
  32. 32 Qwen3-8B Non-thinking 54.4%
  33. 33 deepseek-r1-0528-qwen3-8b Thinking 52.7%
  34. 34 Qwen3-4B Non-thinking 50.3%
  35. 35 Gemma 3 4B Non-thinking 48.3%
  36. 36 yandexgpt-lite Non-thinking 37.6%
  37. 37 Mistral-7B-Instruct-v0.3 Non-thinking 30.1%
  38. 38 Gemma 3 1B Non-thinking 26%
  39. 39 Llama-3.2-3B-Instruct Non-thinking 25.2%
  40. 40 Gemma 3 270M Non-thinking 24.8%
  41. 41 Llama-3.2-1B-Instruct Non-thinking 24.1%

Source: Official PerCoR leaderboard

Source: Official PerCoR leaderboard

#EntryScoreSettingDateSource
1 Gemini 2.5 Pro Jun 2025 94.3% Thinking Jun 2025 Official leaderboard
2 o3 92.2% Thinking Apr 2025 Official leaderboard
3 Claude 3.7 Sonnet 91.2% Non-thinking Feb 2025 Official leaderboard
4 GPT-4.1 88.4% Non-thinking Apr 2025 Official leaderboard
5 Gemini 2.5 Flash Sep 2025 87.2% Thinking Sep 2025 Official leaderboard
6 GPT-4o Mar 2025 86.7% Non-thinking Mar 2025 Official leaderboard
7 GPT-5 mini 86.5% Thinking Aug 2025 Official leaderboard
8 Gemini 2.0 Flash Feb 2025 86.4% Non-thinking Feb 2025 Official leaderboard
9 deepseek-reasoner open weights, 671B 82.5% Thinking Official leaderboard
10 deepseek-chat open weights, 671B 82.4% Non-thinking Official leaderboard
11 Gemini 2.0 Flash-Lite 81.4% Non-thinking Feb 2025 Official leaderboard
12 Cohere Command A 79.9% Non-thinking Mar 2025 Official leaderboard
13 Llama 3.3 70B Instruct 79.6% Non-thinking Dec 2024 Official leaderboard
14 Llama 4 Scout 78.1% Non-thinking Apr 2025 Official leaderboard
15 gpt-oss-120b 77.1% Thinking Aug 2025 Official leaderboard
16 GPT-4.1 mini 77.1% Non-thinking Apr 2025 Official leaderboard
17 Qwen3-32B 76.5% Non-thinking Apr 2025 Official leaderboard
18 Gemma 3 27B 76.3% Non-thinking Mar 2025 Official leaderboard
19 GPT-4o mini 76% Non-thinking Jul 2024 Official leaderboard
20 Command R+ 73.6% Non-thinking Aug 2024 Official leaderboard
21 GPT-5 nano 73.1% Thinking Aug 2025 Official leaderboard
22 Claude 3.5 Haiku 71.6% Non-thinking Oct 2024 Official leaderboard
23 Gemma 3 12B 70.9% Non-thinking Mar 2025 Official leaderboard
24 gpt-oss-20b 69.7% Thinking Aug 2025 Official leaderboard
25 Qwen3-14B 69.6% Non-thinking Apr 2025 Official leaderboard
26 Mistral Small 3.1 68.9% Non-thinking Mar 2025 Official leaderboard
27 Qwen3-30B-A3B 68.8% Non-thinking Apr 2025 Official leaderboard
28 Aya Expanse 32B 63.3% Non-thinking Oct 2024 Official leaderboard
29 Command R (03-2024) 60% Non-thinking Mar 2024 Official leaderboard
30 Gemma 3n E4B 59.1% Non-thinking Jun 2025 Official leaderboard
31 GPT-4.1 nano 54.9% Non-thinking Apr 2025 Official leaderboard
32 Qwen3-8B 54.4% Non-thinking Apr 2025 Official leaderboard
33 deepseek-r1-0528-qwen3-8b 52.7% Thinking May 2025 Official leaderboard
34 Qwen3-4B 50.3% Non-thinking Apr 2025 Official leaderboard
35 Gemma 3 4B 48.3% Non-thinking Mar 2025 Official leaderboard
36 yandexgpt-lite open weights, 8B 37.6% Non-thinking Official leaderboard
37 Mistral-7B-Instruct-v0.3 open weights, 7.3B 30.1% Non-thinking Official leaderboard
38 Gemma 3 1B 26% Non-thinking Mar 2025 Official leaderboard
39 Llama-3.2-3B-Instruct open weights, 3.2B 25.2% Non-thinking Official leaderboard
40 Gemma 3 270M 24.8% Non-thinking Aug 2025 Official leaderboard
41 Llama-3.2-1B-Instruct open weights, 1.2B 24.1% Non-thinking Official leaderboard
#ModelScoreSettingDateSource
1 Gemini 2.5 Pro Jun 2025 94.3% Thinking Jun 2025 Official leaderboard
2 o3 92.2% Thinking Apr 2025 Official leaderboard
3 Claude 3.7 Sonnet 91.2% Non-thinking Feb 2025 Official leaderboard
4 GPT-4.1 88.4% Non-thinking Apr 2025 Official leaderboard
5 Gemini 2.5 Flash Sep 2025 87.2% Thinking Sep 2025 Official leaderboard
6 GPT-4o Mar 2025 86.7% Non-thinking Mar 2025 Official leaderboard
7 GPT-5 mini 86.5% Thinking Aug 2025 Official leaderboard
8 Gemini 2.0 Flash Feb 2025 86.4% Non-thinking Feb 2025 Official leaderboard
9 Gemini 2.0 Flash-Lite 81.4% Non-thinking Feb 2025 Official leaderboard
10 Cohere Command A 79.9% Non-thinking Mar 2025 Official leaderboard
11 Llama 3.3 70B Instruct 79.6% Non-thinking Dec 2024 Official leaderboard
12 Llama 4 Scout 78.1% Non-thinking Apr 2025 Official leaderboard
13 gpt-oss-120b 77.1% Thinking Aug 2025 Official leaderboard
14 GPT-4.1 mini 77.1% Non-thinking Apr 2025 Official leaderboard
15 Qwen3-32B 76.5% Non-thinking Apr 2025 Official leaderboard
16 Gemma 3 27B 76.3% Non-thinking Mar 2025 Official leaderboard
17 GPT-4o mini 76% Non-thinking Jul 2024 Official leaderboard
18 Command R+ 73.6% Non-thinking Aug 2024 Official leaderboard
19 GPT-5 nano 73.1% Thinking Aug 2025 Official leaderboard
20 Claude 3.5 Haiku 71.6% Non-thinking Oct 2024 Official leaderboard
21 Gemma 3 12B 70.9% Non-thinking Mar 2025 Official leaderboard
22 gpt-oss-20b 69.7% Thinking Aug 2025 Official leaderboard
23 Qwen3-14B 69.6% Non-thinking Apr 2025 Official leaderboard
24 Mistral Small 3.1 68.9% Non-thinking Mar 2025 Official leaderboard
25 Qwen3-30B-A3B 68.8% Non-thinking Apr 2025 Official leaderboard
26 Aya Expanse 32B 63.3% Non-thinking Oct 2024 Official leaderboard
27 Command R (03-2024) 60% Non-thinking Mar 2024 Official leaderboard
28 Gemma 3n E4B 59.1% Non-thinking Jun 2025 Official leaderboard
29 GPT-4.1 nano 54.9% Non-thinking Apr 2025 Official leaderboard
30 Qwen3-8B 54.4% Non-thinking Apr 2025 Official leaderboard
31 deepseek-r1-0528-qwen3-8b 52.7% Thinking May 2025 Official leaderboard
32 Qwen3-4B 50.3% Non-thinking Apr 2025 Official leaderboard
33 Gemma 3 4B 48.3% Non-thinking Mar 2025 Official leaderboard
34 Gemma 3 1B 26% Non-thinking Mar 2025 Official leaderboard
35 Gemma 3 270M 24.8% Non-thinking Aug 2025 Official leaderboard

Rankings by part

We read 44 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as acc strict, donyaeeqtesad acc and isna acc. Gemini 2.5 Pro leads 34 of them, and 5 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.

acc strict

35 models we track have a score here.

  1. 1 Gemini 2.5 Pro Jun 2025 94.3%
  2. 2 o3 92.2%
  3. 3 Claude 3.7 Sonnet 91.2%
  4. 4 GPT-4.1 88.4%
  5. 5 Gemini 2.5 Flash Sep 2025 87.1%
  6. 6 GPT-5 mini ↑1 86.5%
  7. 7 Gemini 2.0 Flash Feb 2025 ↑1 86.4%
  8. 8 Gemini 2.0 Flash-Lite ↑1 81.4%
  9. 9 Cohere Command A ↑1 79.8%
  10. 10 GPT-4o Mar 2025 ↓4 78.3%

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
GPT-5 mini
Thinking minimal rea…
86.5%
gpt-oss-120b
Thinking low reasoni…
77.1%
GPT-5 nano
Thinking minimal rea…
73.1%
gpt-oss-20b
Thinking low reasoni…
69.7%

Source: Official PerCoR leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Gemini 2.5 Pro Jun 2025 Thinking 94.3%
o3 Thinking 92.2%
Claude 3.7 Sonnet Non-thinking 91.2%
GPT-4.1 Non-thinking 88.4%
Gemini 2.5 Flash Sep 2025 Thinking 87.2%
GPT-4o Mar 2025 Non-thinking 86.7%
GPT-5 mini Thinking 86.5%
Gemini 2.0 Flash Feb 2025 Non-thinking 86.4%
Gemini 2.0 Flash-Lite Non-thinking 81.4%
Cohere Command A Non-thinking 79.9%
Llama 3.3 70B Instruct Non-thinking 79.6%
Llama 4 Scout Non-thinking 78.1%
gpt-oss-120b Thinking 77.1%
GPT-4.1 mini Non-thinking 77.1%
GPT-5 mini minimal reasoning 76.9%
Qwen3-32B Non-thinking 76.5%
Gemma 3 27B Non-thinking 76.3%
GPT-4o mini Non-thinking 76%
Command R+ Non-thinking 73.6%
GPT-5 nano Thinking 73.1%
Claude 3.5 Haiku Non-thinking 71.6%
gpt-oss-120b low reasoning 71.2%
Gemma 3 12B Non-thinking 70.9%
gpt-oss-20b Thinking 69.7%
Qwen3-14B Non-thinking 69.6%
Mistral Small 3.1 Non-thinking 68.9%
Qwen3-30B-A3B Non-thinking 68.8%
gpt-oss-20b low reasoning 64.5%
Aya Expanse 32B Non-thinking 63.3%
Command R (03-2024) Non-thinking 60%
Gemma 3n E4B Non-thinking 59.1%
GPT-4.1 nano Non-thinking 54.9%
Qwen3-8B Non-thinking 54.4%
deepseek-r1-0528-qwen3-8b Thinking 52.7%
GPT-5 nano minimal reasoning 51.9%
Qwen3-4B Non-thinking 50.3%
Gemma 3 4B Non-thinking 48.3%
Gemma 3 1B Non-thinking 26%
Gemma 3 270M Non-thinking 24.8%

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
GPT-5 mini
Thinking minimal rea…
86.5%
gpt-oss-120b
Thinking low reasoni…
77.1%
GPT-5 nano
Thinking minimal rea…
73.1%
gpt-oss-20b
Thinking low reasoni…
69.7%

Source: Official PerCoR leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Gemini 2.5 Pro Jun 2025 Thinking 94.3%
o3 Thinking 92.2%
Claude 3.7 Sonnet Non-thinking 91.2%
GPT-4.1 Non-thinking 88.4%
Gemini 2.5 Flash Sep 2025 Thinking 87.2%
GPT-4o Mar 2025 Non-thinking 86.7%
GPT-5 mini Thinking 86.5%
Gemini 2.0 Flash Feb 2025 Non-thinking 86.4%
Gemini 2.0 Flash-Lite Non-thinking 81.4%
Cohere Command A Non-thinking 79.9%
Llama 3.3 70B Instruct Non-thinking 79.6%
Llama 4 Scout Non-thinking 78.1%
gpt-oss-120b Thinking 77.1%
GPT-4.1 mini Non-thinking 77.1%
GPT-5 mini minimal reasoning 76.9%
Qwen3-32B Non-thinking 76.5%
Gemma 3 27B Non-thinking 76.3%
GPT-4o mini Non-thinking 76%
Command R+ Non-thinking 73.6%
GPT-5 nano Thinking 73.1%
Claude 3.5 Haiku Non-thinking 71.6%
gpt-oss-120b low reasoning 71.2%
Gemma 3 12B Non-thinking 70.9%
gpt-oss-20b Thinking 69.7%
Qwen3-14B Non-thinking 69.6%
Mistral Small 3.1 Non-thinking 68.9%
Qwen3-30B-A3B Non-thinking 68.8%
gpt-oss-20b low reasoning 64.5%
Aya Expanse 32B Non-thinking 63.3%
Command R (03-2024) Non-thinking 60%
Gemma 3n E4B Non-thinking 59.1%
GPT-4.1 nano Non-thinking 54.9%
Qwen3-8B Non-thinking 54.4%
deepseek-r1-0528-qwen3-8b Thinking 52.7%
GPT-5 nano minimal reasoning 51.9%
Qwen3-4B Non-thinking 50.3%
Gemma 3 4B Non-thinking 48.3%
Gemma 3 1B Non-thinking 26%
Gemma 3 270M Non-thinking 24.8%

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.

OpenAI Google DeepMind
Own leaderboard Passed 85% in May 2025, by o3

Source: Official PerCoR leaderboard

Gemini 2.0 Flash on the official leaderboard passed 85% in February 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
On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2024 2025 Near the ceiling (85%) Gemini 2.5 Pro Gemini 2.0 Flash Llama 3.3 70B Instruct GPT-4o mini Command R (03-2024) 0% 20% 40% 60% 80% 100% 2024 2025 Near the ceiling (85%) Gemini 2.5 Pro Gemini 2.0 Flash

Source: Official PerCoR 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 official leaderboard (41 entries).
  • Pass Models scored 35 models we track, plus 6 other entries on its leaderboard.
  • Note Newest score The newest came 14 months ago, in September 2025, for Gemini 2.5 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 o3 (#84).
  • Note Headroom Saturated. The best is 94.3%, past 85% since February 2025.
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.

How we check