Benchmark Almanac

BhashaBench-Krishi

gpt-4o (snapshot unspecified) 73.5% tops its leaderboard Qwen3-235B-A22B-Instruct Jul 2025 72.7% leads the models we track
Not in model cards
Version V1

Models answer English and Hindi questions from Indian government agricultural exams. Topics include agronomy, horticulture and animal husbandry.

The test contains 15,405 questions. The authors require approval to download it.

Made by
BharatGen
Released
October 2025
Kind
Exam questions
Level
University
Scored
Checked automatically
Answers
Multiple choice
Tasks
All public
Measures
Capability
Language
English, Hindi
In and out
Text → Text
Interaction
Single turn
Items from
Exams and contests
Run by
Anyone
Strongest model tested
gpt-oss-120b #126 on the capability index
Last model added
August 2025
Scores refreshed
November 20th, 2026
31
Entries in its paper, 14 models we track

gpt-4o (snapshot unspecified) tops the leaderboard with 73.5% Qwen3-235B-A22B-Instruct leads with 72.7% among models released by August 2025

Best score per entry from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. 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 Qwen3-235B-A22B-Instruct, at 72.7%. This view shows only models we track, each at its best.

The newest model with a score here came out in August 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 gpt-4o (snapshot unspecified) Zero-shot 73.5%
  2. 2 Qwen3-235B-A22B-Instruct Jul 2025, Zero-shot 72.7%
  3. 3 gpt-oss-120b Zero-shot 69.4%
  4. 4 deepseek-v3 (snapshot unspecified) Zero-shot 59.7%
  5. 5 Gemma 2 27B Zero-shot 58.1%
  6. 6 Gemma 2 9B Zero-shot 53.2%
  7. 7 gemma-2-27b-it Zero-shot 52.5%
  8. 8 gemma-2-9b-it Zero-shot 46.7%
  9. 9 Llama-3.1-8B-Instruct Zero-shot 45%
  10. 10 Pangea-7B Zero-shot 44.9%
  11. 11 gpt-oss-20b Zero-shot 44.7%
  12. 12 Qwen2.5-3B Zero-shot 42.5%
  13. 13 Indic-gemma-7b-finetuned-sft-Navarasa-2.0 Zero-shot 40.7%
  14. 14 Qwen2.5-3B-Instruct Zero-shot 39.9%
  15. 15 gemma-2-2b Zero-shot 38.8%
  16. 16 Llama-3.2-3B-Instruct Zero-shot 38.5%
  17. 17 Llama 3.1-8B Zero-shot 38.1%
  18. 18 Nemotron-4-Mini-Hindi-4B-Base Zero-shot 36.6%
  19. 19 aya-23-8B Zero-shot 36.4%
  20. 20 Granite-3.1-2B-Instruct Zero-shot 36%
  21. 21 Nemotron-4-Mini-Hindi-4B-Instruct Zero-shot 35.7%
  22. 22 Gemma 2 2B IT Zero-shot 34.5%
  23. 23 Param-1 Zero-shot 32.2%
  24. 24 Granite-3.1-3B-A800M-Base Zero-shot 32.2%
  25. 25 Llama 3.2 3B Zero-shot 32%
  26. 26 sarvam-1 Zero-shot 30.2%
  27. 27 Llama 3.2 1B Zero-shot 28.9%
  28. 28 Llama-3.2-1B-Instruct Zero-shot 28.6%
  29. 29 sarvam-2b-v0.5 Zero-shot 27.7%
  30. 30 Gemma 3 270M Zero-shot 27.1%
  31. 31 Gemma 3 270M Base Zero-shot 26.2%

Source: The BhashaBench-Krishi paper

#EntryScoreSettingDateSource
1 gpt-4o (snapshot unspecified) 73.5% Zero-shot Official leaderboard
2 Qwen3-235B-A22B-Instruct Jul 2025 72.7% Zero-shot Jul 2025 Official leaderboard
3 gpt-oss-120b 69.4% Zero-shot Aug 2025 Official leaderboard
4 deepseek-v3 (snapshot unspecified) 59.7% Zero-shot Official leaderboard
5 Gemma 2 27B 58.1% Zero-shot Jun 2024 Official leaderboard
6 Gemma 2 9B 53.2% Zero-shot Jun 2024 Official leaderboard
7 gemma-2-27b-it 52.5% Zero-shot Official leaderboard
8 gemma-2-9b-it 46.7% Zero-shot Official leaderboard
9 Llama-3.1-8B-Instruct 45% Zero-shot Official leaderboard
10 Pangea-7B 44.9% Zero-shot Official leaderboard
11 gpt-oss-20b 44.7% Zero-shot Aug 2025 Official leaderboard
12 Qwen2.5-3B 42.5% Zero-shot Sep 2024 Official leaderboard
13 Indic-gemma-7b-finetuned-sft-Navarasa-2.0 40.7% Zero-shot Official leaderboard
14 Qwen2.5-3B-Instruct 39.9% Zero-shot Official leaderboard
15 gemma-2-2b 38.8% Zero-shot Official leaderboard
16 Llama-3.2-3B-Instruct 38.5% Zero-shot Official leaderboard
17 Llama 3.1-8B 38.1% Zero-shot Jul 2024 Official leaderboard
18 Nemotron-4-Mini-Hindi-4B-Base 36.6% Zero-shot Official leaderboard
19 aya-23-8B 36.4% Zero-shot Official leaderboard
20 Granite-3.1-2B-Instruct 36% Zero-shot Dec 2024 Official leaderboard
21 Nemotron-4-Mini-Hindi-4B-Instruct 35.7% Zero-shot Official leaderboard
22 Gemma 2 2B IT 34.5% Zero-shot Jul 2024 Official leaderboard
23 Param-1 32.2% Zero-shot Official leaderboard
24 Granite-3.1-3B-A800M-Base 32.2% Zero-shot Dec 2024 Official leaderboard
25 Llama 3.2 3B 32% Zero-shot Sep 2024 Official leaderboard
26 sarvam-1 30.2% Zero-shot Official leaderboard
27 Llama 3.2 1B 28.9% Zero-shot Sep 2024 Official leaderboard
28 Llama-3.2-1B-Instruct 28.6% Zero-shot Official leaderboard
29 sarvam-2b-v0.5 27.7% Zero-shot Official leaderboard
30 Gemma 3 270M 27.1% Zero-shot Aug 2025 Official leaderboard
31 Gemma 3 270M Base 26.2% Zero-shot Aug 2025 Official leaderboard
#ModelScoreSettingDateSource
1 Qwen3-235B-A22B-Instruct Jul 2025 72.7% Zero-shot Jul 2025 Official leaderboard
2 gpt-oss-120b 69.4% Zero-shot Aug 2025 Official leaderboard
3 Gemma 2 27B 58.1% Zero-shot Jun 2024 Official leaderboard
4 Gemma 2 9B 53.2% Zero-shot Jun 2024 Official leaderboard
5 gpt-oss-20b 44.7% Zero-shot Aug 2025 Official leaderboard
6 Qwen2.5-3B 42.5% Zero-shot Sep 2024 Official leaderboard
7 Llama 3.1-8B 38.1% Zero-shot Jul 2024 Official leaderboard
8 Granite-3.1-2B-Instruct 36% Zero-shot Dec 2024 Official leaderboard
9 Gemma 2 2B IT 34.5% Zero-shot Jul 2024 Official leaderboard
10 Granite-3.1-3B-A800M-Base 32.2% Zero-shot Dec 2024 Official leaderboard
11 Llama 3.2 3B 32% Zero-shot Sep 2024 Official leaderboard
12 Llama 3.2 1B 28.9% Zero-shot Sep 2024 Official leaderboard
13 Gemma 3 270M 27.1% Zero-shot Aug 2025 Official leaderboard
14 Gemma 3 270M Base 26.2% Zero-shot Aug 2025 Official leaderboard

Rankings by part

We read 10 parts from the benchmark’s paper, each ranking models on one part of the test, such as English, Hindi and Easy. Qwen3-235B-A22B-Instruct leads every one of them. Arrows show how far a model moved from its overall rank among the models we track.

English

14 models we track have a score here.

  1. 1 Qwen3-235B-A22B-Instruct Jul 2025 74.6%
  2. 2 gpt-oss-120b 71.4%
  3. 3 Gemma 2 27B 59.8%
  4. 4 Gemma 2 9B 55.2%
  5. 5 gpt-oss-20b 46.6%
  6. 6 Qwen2.5-3B 44.6%
  7. 7 Llama 3.1-8B 39.5%
  8. 8 Granite-3.1-2B-Instruct 37.7%
  9. 9 Gemma 2 2B IT 35.9%
  10. 10 Granite-3.1-3B-A800M-Base 33.4%

Number one over time

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

Qwen
Benchmark paper Not tested since Oct 2025

Source: The BhashaBench-Krishi paper

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
In the benchmark’s paper Best so far
0% 20% 40% 60% 80% 100% 2024 2025 Near the ceiling (85%) Qwen3-235B-A22B-Instruct Gemma 2 27B 0% 20% 40% 60% 80% 100% 2024 2025 Near the ceiling (85%) Qwen3-235B-A22B-Instruct Gemma 2 27B

Source: Official BhashaBench-Krishi 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.

Versions

More in Knowledge

All 333 in Knowledge
  • MMLU-Pro In use Harder 10-option version of MMLU with reasoning-focused questions across 14 subjects. Intern-S2-397B 89.8%
  • MMLU In use 57-subject multiple-choice exam questions from elementary to professional level. Scored by accuracy. Seed1.8 92.3%
  • SuperGPQA In use 26,529 graduate-level multiple-choice questions across 285 disciplines. Accuracy. Qwen 3.6 Max (Preview) 73.9%
  • WideSearch In use 200 bilingual broad info-seeking tasks: agent collects many atomic facts into a table. Scored by success rate and row/item F1. Tencent Hy4 preview 83.9%
  • BBEH micro average In use Models solve harder replacements for each of the 23 BIG-Bench Hard tasks, 4,520 examples in total. The score is the share of all examples answered correctly. Gemma 4 31B IT 74.4%
  • HLE-Diamond In use Models answer 1,000 cleaned expert questions from Humanity's Last Exam, 500 on reasoning and 500 on knowledge, closed book. Graded on exact answers. Claude Fable 5.2 77.9%

Data audit

All 9 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 3 pass 2 notes
  • Pass Independent runs Scores from the benchmark’s paper (31 entries).
  • Pass Models scored 14 models we track, plus 17 other entries on its leaderboard.
  • Note Newest score The newest came 15 months ago, in August 2025, for Gemma 3 270M. That’s expected for a test replaced by BhashaBench-Multi (Krishi).
  • Note Frontier models None of the top 10 has a score, which is fine for a retired test. The strongest tested is gpt-oss-120b (#126).
  • Pass Headroom The best is 72.7%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on arxiv.org: 14 rows for models we track and 17 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