Benchmark Almanac

EnviroExam

deepseek-67b-chat 84.9 tops its leaderboard Qwen1.5-7B-Chat 80.6 leads the models we track
Not in model cards

Models answer multiple choice questions from 42 environmental science courses. The composite index discounts mean accuracy when performance varies across courses.

The EnvGPT team evaluated five models on EnviroExam. The original release has 210 development questions and 726 test questions. Environmental biotechnology accounts for 20 questions across the two splits.

Released
May 2024
Kind
Science questions
Level
University
Scored
Checked automatically
Answers
Multiple choice
Tasks
All public
Measures
Capability
Language
English
In and out
Text → Text
Interaction
Single turn
Items from
Exams and contests
Refresh
One fixed set
Run by
Anyone
Strongest model tested
Gemma 7B #211 on the capability index
Last model added
February 2024
Scores refreshed
November 20th, 2026
31
Entries in its paper, 10 models we track

deepseek-67b-chat tops the leaderboard with 84.9 Qwen1.5-7B-Chat leads with 80.6 among models released by February 2024

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. 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. The best model we track is Qwen1.5-7B-Chat, at 80.6. This view shows only models we track, each at its best.

The newest model with a score here came out in February 2024, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 deepseek-67b-chat Five-shot 84.9
  2. 2 llama-3-70b-instruct Five-shot 84.5
  3. 3 mixtral-8x7b-instruct-v0.1 Five-shot 80.8
  4. 4 Qwen1.5-7B-Chat Five-shot 80.6
  5. 5 internlm2-chat-7b Five-shot 79.8
  6. 6 llama-3-8b-instruct Five-shot 78.6
  7. 7 internlm2-chat-20b Five-shot 78.2
  8. 8 internlm-chat-20b Five-shot 75.6
  9. 9 Qwen1.5-14B-Chat Five-shot 75.6
  10. 10 Yi-6B-Chat Five-shot 75.1
  11. 11 Qwen-14B Five-shot 74.7
  12. 12 Baichuan2-13B-Chat Zero-shot 73.2
  13. 13 chatglm3-6b-32k Five-shot 71.3
  14. 14 qwen-7b-chat Five-shot 69.8
  15. 15 bluelm-7b-chat Zero-shot 68.1
  16. 16 ChatGLM3-6B Five-shot 67.4
  17. 17 internlm-chat-7b Zero-shot 67.3
  18. 18 baichuan-13b-chat Zero-shot 67.1
  19. 19 llama-2-70b-chat Five-shot 66.7
  20. 20 deepseek-7b-chat Five-shot 63.1
  21. 21 Gemma 7B Zero-shot 59.4
  22. 22 tigerbot-13b-chat-v2 Zero-shot 59.2
  23. 23 vicuna-13b-v1.5 Five-shot 59.1
  24. 24 vicuna-7b-v1.5-16k Five-shot 53.3
  25. 25 tigerbot-7b-chat-v3 Zero-shot 51.2
  26. 26 Mistral-7B-Instruct-v0.1 Five-shot 49.1
  27. 27 llama-2-13b-chat Five-shot 47.8
  28. 28 gemma-2b-it Zero-shot 47.3
  29. 29 chatglm2-6b Zero-shot 46.0
  30. 30 skywork-13b Zero-shot 17.8
  31. 31 vicuna-7b-v1.5 Zero-shot 15.4

Source: The EnviroExam paper

#EntryScoreSettingDateSource
1 deepseek-67b-chat 84.9 Five-shot Official leaderboard
2 llama-3-70b-instruct 84.5 Five-shot Official leaderboard
3 mixtral-8x7b-instruct-v0.1 80.8 Five-shot Official leaderboard
4 Qwen1.5-7B-Chat 80.6 Five-shot Feb 2024 Official leaderboard
5 internlm2-chat-7b 79.8 Five-shot Official leaderboard
6 llama-3-8b-instruct 78.6 Five-shot Official leaderboard
7 internlm2-chat-20b 78.2 Five-shot Official leaderboard
8 internlm-chat-20b 75.6 Five-shot Sep 2023 Official leaderboard
9 Qwen1.5-14B-Chat 75.6 Five-shot Feb 2024 Official leaderboard
10 Yi-6B-Chat 75.1 Five-shot Nov 2023 Official leaderboard
11 Qwen-14B 74.7 Five-shot Sep 2023 Official leaderboard
12 Baichuan2-13B-Chat 73.2 Zero-shot Sep 2023 Official leaderboard
13 chatglm3-6b-32k 71.3 Five-shot Official leaderboard
14 qwen-7b-chat 69.8 Five-shot Official leaderboard
15 bluelm-7b-chat 68.1 Zero-shot Official leaderboard
16 ChatGLM3-6B 67.4 Five-shot Oct 2023 Official leaderboard
17 internlm-chat-7b 67.3 Zero-shot Official leaderboard
18 baichuan-13b-chat 67.1 Zero-shot Official leaderboard
19 llama-2-70b-chat 66.7 Five-shot Official leaderboard
20 deepseek-7b-chat 63.1 Five-shot Official leaderboard
21 Gemma 7B 59.4 Zero-shot Feb 2024 Official leaderboard
22 tigerbot-13b-chat-v2 59.2 Zero-shot Official leaderboard
23 vicuna-13b-v1.5 59.1 Five-shot Official leaderboard
24 vicuna-7b-v1.5-16k 53.3 Five-shot Official leaderboard
25 tigerbot-7b-chat-v3 51.2 Zero-shot Official leaderboard
26 Mistral-7B-Instruct-v0.1 49.1 Five-shot Sep 2023 Official leaderboard
27 llama-2-13b-chat 47.8 Five-shot Official leaderboard
28 gemma-2b-it 47.3 Zero-shot Official leaderboard
29 chatglm2-6b 46.0 Zero-shot Jun 2023 Official leaderboard
30 skywork-13b 17.8 Zero-shot Official leaderboard
31 vicuna-7b-v1.5 15.4 Zero-shot Official leaderboard
#ModelScoreSettingDateSource
1 Qwen1.5-7B-Chat 80.6 Five-shot Feb 2024 Official leaderboard
2 internlm-chat-20b 75.6 Five-shot Sep 2023 Official leaderboard
3 Qwen1.5-14B-Chat 75.6 Five-shot Feb 2024 Official leaderboard
4 Yi-6B-Chat 75.1 Five-shot Nov 2023 Official leaderboard
5 Qwen-14B 74.7 Five-shot Sep 2023 Official leaderboard
6 Baichuan2-13B-Chat 73.2 Zero-shot Sep 2023 Official leaderboard
7 ChatGLM3-6B 67.4 Five-shot Oct 2023 Official leaderboard
8 Gemma 7B 59.4 Zero-shot Feb 2024 Official leaderboard
9 Mistral-7B-Instruct-v0.1 49.1 Five-shot Sep 2023 Official leaderboard
10 chatglm2-6b 46.0 Zero-shot Jun 2023 Official leaderboard

Rankings by part

We read 43 parts from the benchmark’s paper, each ranking models on one part of the test, such as environmental science, environmental fluid mechanics and environmental analytical chemistry. Baichuan2-13B-Chat leads 9 of them, and 8 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.

environmental science

10 models we track have a score here.

  1. 1 Qwen1.5-14B-Chat ↑2 100%
  2. 2 Gemma 7B ↑6 100%
  3. 3 Baichuan2-13B-Chat ↑3 94.4%
  4. 4 internlm-chat-20b ↓2 94.4%
  5. 5 Qwen-14B 94.4%
  6. 6 Qwen1.5-7B-Chat ↓5 94.4%
  7. 7 Yi-6B-Chat ↓3 88.9%
  8. 8 Mistral-7B-Instruct-v0.1 ↑1 83.3%
  9. 9 ChatGLM3-6B ↓2 77.8%
  10. 10 chatglm2-6b 55.6%

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
Qwen1.5-7B-Chat
Zero-shot Five-shot
80.6
internlm-chat-20b
Zero-shot
75.6
Qwen1.5-14B-Chat
Zero-shot
75.6
Yi-6B-Chat
Zero-shot
75.1
Qwen-14B
Zero-shot Five-shot
74.7
Baichuan2-13B-Chat
Five-shot
73.2
ChatGLM3-6B
Zero-shot
67.4
Gemma 7B
Five-shot Zero-shot
59.4
Mistral-7B-Instruct-v0.1
Zero-shot
49.1
chatglm2-6b
Five-shot Zero-shot
46.0

Source: The EnviroExam paper

ModelEffortHarnessScoreCost per taskOutput tokens
Qwen1.5-7B-Chat Five-shot 80.6
internlm-chat-20b Five-shot 75.6
Qwen1.5-14B-Chat Five-shot 75.6
Yi-6B-Chat Five-shot 75.1
Qwen-14B Five-shot 74.7
Baichuan2-13B-Chat Zero-shot 73.2
Yi-6B-Chat Zero-shot 71.0
Qwen1.5-14B-Chat Zero-shot 70.7
internlm-chat-20b Zero-shot 69.4
ChatGLM3-6B Five-shot 67.4
Baichuan2-13B-Chat Five-shot 66.6
Qwen-14B Zero-shot 65.3
Qwen1.5-7B-Chat Zero-shot 65.1
ChatGLM3-6B Zero-shot 62.7
Gemma 7B Zero-shot 59.4
Mistral-7B-Instruct-v0.1 Five-shot 49.1
Mistral-7B-Instruct-v0.1 Zero-shot 47.2
chatglm2-6b Zero-shot 46.0
Gemma 7B Five-shot 31.5
chatglm2-6b Five-shot 3.1

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.

Lowest effort to highest, light to dark
Qwen1.5-7B-Chat
Zero-shot Five-shot
80.6
internlm-chat-20b
Zero-shot
75.6
Qwen1.5-14B-Chat
Zero-shot
75.6
Yi-6B-Chat
Zero-shot
75.1
Qwen-14B
Zero-shot Five-shot
74.7
Baichuan2-13B-Chat
Five-shot
73.2
ChatGLM3-6B
Zero-shot
67.4
Gemma 7B
Five-shot Zero-shot
59.4
Mistral-7B-Instruct-v0.1
Zero-shot
49.1
chatglm2-6b
Five-shot Zero-shot
46.0

Source: The EnviroExam paper

ModelEffortHarnessScoreCost per taskOutput tokens
Qwen1.5-7B-Chat Five-shot 80.6
internlm-chat-20b Five-shot 75.6
Qwen1.5-14B-Chat Five-shot 75.6
Yi-6B-Chat Five-shot 75.1
Qwen-14B Five-shot 74.7
Baichuan2-13B-Chat Zero-shot 73.2
Yi-6B-Chat Zero-shot 71.0
Qwen1.5-14B-Chat Zero-shot 70.7
internlm-chat-20b Zero-shot 69.4
ChatGLM3-6B Five-shot 67.4
Baichuan2-13B-Chat Five-shot 66.6
Qwen-14B Zero-shot 65.3
Qwen1.5-7B-Chat Zero-shot 65.1
ChatGLM3-6B Zero-shot 62.7
Gemma 7B Zero-shot 59.4
Mistral-7B-Instruct-v0.1 Five-shot 49.1
Mistral-7B-Instruct-v0.1 Zero-shot 47.2
chatglm2-6b Zero-shot 46.0
Gemma 7B Five-shot 31.5
chatglm2-6b Five-shot 3.1

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 May 2024

Source: The EnviroExam 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
0 20 40 60 80 100 2023 2024 0 20 40 60 80 100 2023 2024

Source: Official EnviroExam 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 Science and reasoning

All 203 in Science and reasoning
  • GPQA Diamond In use 198 graduate-level 'Google-proof' multiple-choice questions in biology, chemistry and physics. Gemini 4 Carbon 97%
  • Terminal-Bench-Science In use 70 scientific research-workflow tasks across five science areas, run in Harbor containers. Hidden tests give the resolution rate. GPT-6 Astra 68.1%
  • BioMysteryBench In use Bioinformatics problems on anonymized real datasets that require real analysis with tools. Rubric-graded, with human-solvable and human-difficult splits. Claude Fable 5.2 83.7%
  • FrontierScience In use Models solve physics, chemistry, and biology problems in an Olympiad track graded by short answer and a Research track graded by 10-point rubrics. Intern-S2-397B 87%
  • HiPhO In use Models solve theory problems from 2024 and 2025 physics olympiad exams, some with diagrams. Official marking schemes grade answers and steps, mapped to medal levels. Gemini 3 Pro 78.3%
  • LABBench2 In use 11 biology research sub-tasks (literature, figures, tables, patents, trials and more) with tool access. Macro-averaged accuracy. GPT-5.2 60.9%

Data audit

8 of 9 checks pass, 1 warning. Look first at the Newest score check. The newest came 2 years ago, in February 2024, for Gemma 7B. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 2 pass 2 notes
  • Pass Independent runs Scores from the benchmark’s paper (31 entries).
  • Pass Models scored 10 models we track, plus 21 other entries on its leaderboard.
  • Warning Newest score The newest came 2 years ago, in February 2024, for Gemma 7B.
  • Note Frontier models None of the top 10 has a score. The strongest tested is Gemma 7B (#211).
  • Note Headroom Scored as an index, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on huggingface.co: 20 rows for models we track and 21 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