Benchmark Almanac

C-Eval

Qwen3-Max 93.7% leads since January 2026
Claude 1.3 54.2% leads since May 2023
Qwen3-Max 93.7% leads since January 2026
Fading

13,948 Chinese multiple-choice exam questions across 52 disciplines and four difficulty levels. Accuracy.

Saturated. The maintainers released the full test set with answers on July 27th, 2025, so test scores are no longer held out.

Made by
SJTU + Tsinghua + Edinburgh
Released
May 2023
Kind
Exam questions
Level
University
Scored
Checked automatically
Answers
Multiple choice
Tasks
All public
Strongest model tested
DeepSeek V4.1 Flash #34 on the capability index
Last in a launch post
September 2026, dropped after it saturated
Last model added
September 2026
Scores refreshed
November 12th, 2026

In the authors’ words

New NLP benchmarks are urgently needed to align with the rapid development of large language models (LLMs). We present C-Eval, the first comprehensive Chinese evaluation suite designed to assess advanced knowledge and reasoning abilities of foundation models in a Chinese context. C-Eval comprises multiple-choice questions across four difficulty levels: middle school, high school, college, and professional. The questions span 52 diverse disciplines, ranging from humanities to science and engineering. C-Eval is accompanied by C-Eval Hard, a subset of very challenging subjects in C-Eval that requires advanced reasoning abilities to solve. We conduct a comprehensive evaluation of the most advanced LLMs on C-Eval, including both English- and Chinese-oriented models. Results indicate that only GPT-4 could achieve an average accuracy of over 60%, suggesting that there is still significant room for improvement for current LLMs. We anticipate C-Eval will help analyze important strengths and shortcomings of foundation models, and foster their development and growth for Chinese users.

Abstract of C-Eval: A Multi-Level Multi-Discipline Chinese Evaluation Suite for Foundation Models (arXiv:2305.08322, CC0)
46
Models scored in lab cards
36
Launch posts list it, from 9 labs
26
System cards report it, from 9 labs

Qwen3-Max leads with 93.7% among models released by September 2026

Best score per model as labs report them in their own cards. Labs pick their own settings, so these compare less cleanly than independent runs.

Claude 1.3 leads with 54.2% across all independent runs

Best score per model across the official leaderboard, Epoch AI’s own runs and any board Epoch AI copied. Each bar and row names its source.

Qwen3-Max leads with 93.7%, 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 September 2026, and nobody has reported a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

46 models. Scroll sideways for the rest.

  1. 93.7 Qwen3-Max
  2. 93.3 Qwen 3.6 Plus
  3. 93.1 DeepSeek-V4-Pro
  4. 93.0 Qwen3.5 397B-A17B
  5. 92.1 DeepSeek V4.1 Flash
  6. 92.1 DeepSeek-V4.1-Flash Base
  7. 92.0 Qwen3.5-Omni-Plus
  8. 91.9 Qwen3.5-122B-A10B
  9. 91.8 DeepSeek-R1
  10. 91.8 Hunyuan-T1
  11. 91.7 LongCat-Flash-Omni
  12. 91.5 Ring-1T
  13. 91.4 Qwen3.6 27B
  14. 90.9 Hunyuan-TurboS
  15. 90.9 InternVL3.5-241B-A28B
  16. 90.6 ERNIE-4.5-300B-A47B
  17. 90.5 Qwen3.5-27B
  18. 90.4 LongCat-Flash-Chat
  19. 90.2 Qwen3.5-35B-A3B
  20. 90.0 Qwen 3.6 35B-A3B
  21. 89.6 Qwen3-235B-A22B
  22. 89.2 Keye-VL-2.0-30B-A3B
  23. 88.8 GLM-5
  24. 88.3 Kimi k1.5
  25. 88.2 Qwen3-Coder-Next
  26. 88.2 Qwen3.5-9B
  27. 87.3 Qwen3-32B
  28. 86.9 GLM-4.5
  29. 86.6 Qwen3-30B-A3B
  30. 86.5 DeepSeek-V3
  31. 86.2 Qwen3-14B
  32. 85.1 Qwen3.5-4B
  33. 84.1 Qwen1.5-72B
  34. 83.8 Qwen2-72B
  35. 83.4 Qwen3-8B
  36. 78.0 DeepSeek-V2 (MoE-236B, May 2024)
  37. 77.5 Qwen3-4B
  38. 77.2 Moonlight-16B-A3B
  39. 73.2 Qwen3.5-2B
  40. 71.7 Qwen-14B
  41. 69.0 ChatGLM3-6B
  42. 68.1 Qwen3-1.7B
  43. 65.2 DeepSeek LLM 67B
  44. 59.7 Qwen-7B
  45. 50.7 Qwen-1_8B
  46. 50.5 Qwen3.5-0.8B
  1. 1 Qwen3-Max 93.7%
  2. 2 Qwen 3.6 Plus 93.3%
  3. 3 DeepSeek-V4-Pro 93.1%
  4. 4 Qwen3.5 397B-A17B 93%
  5. 5 DeepSeek V4.1 Flash 92.1%
  6. 6 DeepSeek-V4.1-Flash Base 92.1%
  7. 7 Qwen3.5-Omni-Plus 92%
  8. 8 Qwen3.5-122B-A10B 91.9%
  9. 9 DeepSeek-R1 91.8%
  10. 10 Hunyuan-T1 91.8%
  11. 11 LongCat-Flash-Omni 91.7%
  12. 12 Ring-1T 91.5%
  13. 13 Qwen3.6 27B 91.4%
  14. 14 Hunyuan-TurboS 90.9%
  15. 15 InternVL3.5-241B-A28B 90.9%
  16. 16 ERNIE-4.5-300B-A47B 90.6%
  17. 17 Qwen3.5-27B 90.5%
  18. 18 LongCat-Flash-Chat 90.4%
  19. 19 Qwen3.5-35B-A3B 90.2%
  20. 20 Qwen 3.6 35B-A3B 90%
  21. 21 Qwen3-235B-A22B 89.6%
  22. 22 Keye-VL-2.0-30B-A3B 89.2%
  23. 23 GLM-5 88.8%
  24. 24 Kimi k1.5 88.3%
  25. 25 Qwen3-Coder-Next 88.2%
  26. 26 Qwen3.5-9B 88.2%
  27. 27 Qwen3-32B 87.3%
  28. 28 GLM-4.5 86.9%
  29. 29 Qwen3-30B-A3B 86.6%
  30. 30 DeepSeek-V3 86.5%
  31. 31 Qwen3-14B 86.2%
  32. 32 Qwen3.5-4B 85.1%
  33. 33 Qwen1.5-72B 84.1%
  34. 34 Qwen2-72B 83.8%
  35. 35 Qwen3-8B 83.4%
  36. 36 DeepSeek-V2 (MoE-236B, May 2024) 78%
  37. 37 Qwen3-4B 77.5%
  38. 38 Moonlight-16B-A3B 77.2%
  39. 39 Qwen3.5-2B 73.2%
  40. 40 Qwen-14B 71.7%
  41. 41 ChatGLM3-6B 69%
  42. 42 Qwen3-1.7B 68.1%
  43. 43 DeepSeek LLM 67B 65.2%
  44. 44 Qwen-7B 59.7%
  45. 45 Qwen-1_8B 50.7%
  46. 46 Qwen3.5-0.8B 50.5%

Source: Labs’ model cards

Source: Official C-Eval leaderboard

49 models. Scroll sideways for the rest.

  1. 93.7 Qwen3-Max
  2. 93.3 Qwen 3.6 Plus
  3. 93.1 DeepSeek-V4-Pro
  4. 93.0 Qwen3.5 397B-A17B
  5. 92.1 DeepSeek V4.1 Flash
  6. 92.1 DeepSeek-V4.1-Flash Base
  7. 92.0 Qwen3.5-Omni-Plus
  8. 91.9 Qwen3.5-122B-A10B
  9. 91.8 DeepSeek-R1
  10. 91.8 Hunyuan-T1
  11. 91.7 LongCat-Flash-Omni
  12. 91.5 Ring-1T
  13. 91.4 Qwen3.6 27B
  14. 90.9 Hunyuan-TurboS
  15. 90.9 InternVL3.5-241B-A28B
  16. 90.6 ERNIE-4.5-300B-A47B
  17. 90.5 Qwen3.5-27B
  18. 90.4 LongCat-Flash-Chat
  19. 90.2 Qwen3.5-35B-A3B
  20. 90.0 Qwen 3.6 35B-A3B
  21. 89.6 Qwen3-235B-A22B
  22. 89.2 Keye-VL-2.0-30B-A3B
  23. 88.8 GLM-5
  24. 88.3 Kimi k1.5
  25. 88.2 Qwen3-Coder-Next
  26. 88.2 Qwen3.5-9B
  27. 87.3 Qwen3-32B
  28. 86.9 GLM-4.5
  29. 86.6 Qwen3-30B-A3B
  30. 86.5 DeepSeek-V3
  31. 86.2 Qwen3-14B
  32. 85.1 Qwen3.5-4B
  33. 84.1 Qwen1.5-72B
  34. 83.8 Qwen2-72B
  35. 83.4 Qwen3-8B
  36. 78.0 DeepSeek-V2 (MoE-236B, May 2024)
  37. 77.5 Qwen3-4B
  38. 77.2 Moonlight-16B-A3B
  39. 73.2 Qwen3.5-2B
  40. 71.7 Qwen-14B
  41. 69.0 ChatGLM3-6B
  42. 68.1 Qwen3-1.7B
  43. 65.2 DeepSeek LLM 67B
  44. 59.7 Qwen-7B
  45. 54.2 Claude 1.3
  46. 50.7 Qwen-1_8B
  47. 50.5 Qwen3.5-0.8B
  48. 38.9 ChatGLM-6B
  49. 38.8 LLaMA-65B
  1. 1 Qwen3-Max 93.7%
  2. 2 Qwen 3.6 Plus 93.3%
  3. 3 DeepSeek-V4-Pro 93.1%
  4. 4 Qwen3.5 397B-A17B 93%
  5. 5 DeepSeek V4.1 Flash 92.1%
  6. 6 DeepSeek-V4.1-Flash Base 92.1%
  7. 7 Qwen3.5-Omni-Plus 92%
  8. 8 Qwen3.5-122B-A10B 91.9%
  9. 9 DeepSeek-R1 91.8%
  10. 10 Hunyuan-T1 91.8%
  11. 11 LongCat-Flash-Omni 91.7%
  12. 12 Ring-1T 91.5%
  13. 13 Qwen3.6 27B 91.4%
  14. 14 Hunyuan-TurboS 90.9%
  15. 15 InternVL3.5-241B-A28B 90.9%
  16. 16 ERNIE-4.5-300B-A47B 90.6%
  17. 17 Qwen3.5-27B 90.5%
  18. 18 LongCat-Flash-Chat 90.4%
  19. 19 Qwen3.5-35B-A3B 90.2%
  20. 20 Qwen 3.6 35B-A3B 90%
  21. 21 Qwen3-235B-A22B 89.6%
  22. 22 Keye-VL-2.0-30B-A3B 89.2%
  23. 23 GLM-5 88.8%
  24. 24 Kimi k1.5 88.3%
  25. 25 Qwen3-Coder-Next 88.2%
  26. 26 Qwen3.5-9B 88.2%
  27. 27 Qwen3-32B 87.3%
  28. 28 GLM-4.5 86.9%
  29. 29 Qwen3-30B-A3B 86.6%
  30. 30 DeepSeek-V3 86.5%
  31. 31 Qwen3-14B 86.2%
  32. 32 Qwen3.5-4B 85.1%
  33. 33 Qwen1.5-72B 84.1%
  34. 34 Qwen2-72B 83.8%
  35. 35 Qwen3-8B 83.4%
  36. 36 DeepSeek-V2 (MoE-236B, May 2024) 78%
  37. 37 Qwen3-4B 77.5%
  38. 38 Moonlight-16B-A3B 77.2%
  39. 39 Qwen3.5-2B 73.2%
  40. 40 Qwen-14B 71.7%
  41. 41 ChatGLM3-6B 69%
  42. 42 Qwen3-1.7B 68.1%
  43. 43 DeepSeek LLM 67B 65.2%
  44. 44 Qwen-7B 59.7%
  45. 45 Claude 1.3 Five-shot 54.2%
  46. 46 Qwen-1_8B 50.7%
  47. 47 Qwen3.5-0.8B 50.5%
  48. 48 ChatGLM-6B Zero-shot 38.9%
  49. 49 LLaMA-65B Five-shot 38.8%

Sources: Labs’ model cards and official C-Eval leaderboard

#ModelScoreSettingDateSource
1 Qwen3-Max 93.7% Jan 2026 Lab model card
2 Qwen 3.6 Plus 93.3% Apr 2026 Lab model card
3 DeepSeek-V4-Pro 93.1% Apr 2026 Lab model card
4 Qwen3.5 397B-A17B 93% Feb 2026 Lab model card
5 DeepSeek V4.1 Flash 92.1% Sep 2026 Lab model card
6 DeepSeek-V4.1-Flash Base 92.1% Sep 2026 Lab model card
7 Qwen3.5-Omni-Plus 92% Mar 2026 Lab model card
8 Qwen3.5-122B-A10B 91.9% Feb 2026 Lab model card
9 DeepSeek-R1 91.8% Jan 2025 Lab model card
10 Hunyuan-T1 91.8% Mar 2025 Lab model card
11 LongCat-Flash-Omni 91.7% Oct 2025 Lab model card
12 Ring-1T 91.5% Oct 2025 Lab model card
13 Qwen3.6 27B 91.4% Apr 2026 Lab model card
14 Hunyuan-TurboS 90.9% Feb 2025 Lab model card
15 InternVL3.5-241B-A28B 90.9% Aug 2025 Lab model card
16 ERNIE-4.5-300B-A47B 90.6% Jun 2025 Lab model card
17 Qwen3.5-27B 90.5% Feb 2026 Lab model card
18 LongCat-Flash-Chat 90.4% Aug 2025 Lab model card
19 Qwen3.5-35B-A3B 90.2% Feb 2026 Lab model card
20 Qwen 3.6 35B-A3B 90% Apr 2026 Lab model card
21 Qwen3-235B-A22B 89.6% Apr 2025 Lab model card
22 Keye-VL-2.0-30B-A3B 89.2% May 2026 Lab model card
23 GLM-5 88.8% Feb 2026 Lab model card
24 Kimi k1.5 88.3% Jan 2025 Lab model card
25 Qwen3-Coder-Next 88.2% Feb 2026 Lab model card
26 Qwen3.5-9B 88.2% Mar 2026 Lab model card
27 Qwen3-32B 87.3% May 2025 Lab model card
28 GLM-4.5 86.9% Aug 2025 Lab model card
29 Qwen3-30B-A3B 86.6% May 2025 Lab model card
30 DeepSeek-V3 86.5% Dec 2024 Lab model card
31 Qwen3-14B 86.2% May 2025 Lab model card
32 Qwen3.5-4B 85.1% Mar 2026 Lab model card
33 Qwen1.5-72B 84.1% Feb 2024 Lab model card
34 Qwen2-72B 83.8% Jun 2024 Lab model card
35 Qwen3-8B 83.4% May 2025 Lab model card
36 DeepSeek-V2 (MoE-236B, May 2024) 78% May 2024 Lab model card
37 Qwen3-4B 77.5% May 2025 Lab model card
38 Moonlight-16B-A3B 77.2% Feb 2025 Lab model card
39 Qwen3.5-2B 73.2% Mar 2026 Lab model card
40 Qwen-14B 71.7% Sep 2023 Lab model card
41 ChatGLM3-6B 69% Oct 2023 Lab model card
42 Qwen3-1.7B 68.1% May 2025 Lab model card
43 DeepSeek LLM 67B 65.2% Nov 2023 Lab model card
44 Qwen-7B 59.7% Sep 2023 Lab model card
45 Qwen-1_8B 50.7% Sep 2023 Lab model card
46 Qwen3.5-0.8B 50.5% Mar 2026 Lab model card
#ModelScoreSettingDateSource
1 Claude 1.3 54.2% Five-shot Apr 2023 Official leaderboard
2 ChatGLM-6B 38.9% Zero-shot Mar 2023 Official leaderboard
3 LLaMA-65B 38.8% Five-shot Feb 2023 Official leaderboard
#ModelScoreSettingDateSource
1 Qwen3-Max 93.7% Jan 2026 Lab model card
2 Qwen 3.6 Plus 93.3% Apr 2026 Lab model card
3 DeepSeek-V4-Pro 93.1% Apr 2026 Lab model card
4 Qwen3.5 397B-A17B 93% Feb 2026 Lab model card
5 DeepSeek V4.1 Flash 92.1% Sep 2026 Lab model card
6 DeepSeek-V4.1-Flash Base 92.1% Sep 2026 Lab model card
7 Qwen3.5-Omni-Plus 92% Mar 2026 Lab model card
8 Qwen3.5-122B-A10B 91.9% Feb 2026 Lab model card
9 DeepSeek-R1 91.8% Jan 2025 Lab model card
10 Hunyuan-T1 91.8% Mar 2025 Lab model card
11 LongCat-Flash-Omni 91.7% Oct 2025 Lab model card
12 Ring-1T 91.5% Oct 2025 Lab model card
13 Qwen3.6 27B 91.4% Apr 2026 Lab model card
14 Hunyuan-TurboS 90.9% Feb 2025 Lab model card
15 InternVL3.5-241B-A28B 90.9% Aug 2025 Lab model card
16 ERNIE-4.5-300B-A47B 90.6% Jun 2025 Lab model card
17 Qwen3.5-27B 90.5% Feb 2026 Lab model card
18 LongCat-Flash-Chat 90.4% Aug 2025 Lab model card
19 Qwen3.5-35B-A3B 90.2% Feb 2026 Lab model card
20 Qwen 3.6 35B-A3B 90% Apr 2026 Lab model card
21 Qwen3-235B-A22B 89.6% Apr 2025 Lab model card
22 Keye-VL-2.0-30B-A3B 89.2% May 2026 Lab model card
23 GLM-5 88.8% Feb 2026 Lab model card
24 Kimi k1.5 88.3% Jan 2025 Lab model card
25 Qwen3-Coder-Next 88.2% Feb 2026 Lab model card
26 Qwen3.5-9B 88.2% Mar 2026 Lab model card
27 Qwen3-32B 87.3% May 2025 Lab model card
28 GLM-4.5 86.9% Aug 2025 Lab model card
29 Qwen3-30B-A3B 86.6% May 2025 Lab model card
30 DeepSeek-V3 86.5% Dec 2024 Lab model card
31 Qwen3-14B 86.2% May 2025 Lab model card
32 Qwen3.5-4B 85.1% Mar 2026 Lab model card
33 Qwen1.5-72B 84.1% Feb 2024 Lab model card
34 Qwen2-72B 83.8% Jun 2024 Lab model card
35 Qwen3-8B 83.4% May 2025 Lab model card
36 DeepSeek-V2 (MoE-236B, May 2024) 78% May 2024 Lab model card
37 Qwen3-4B 77.5% May 2025 Lab model card
38 Moonlight-16B-A3B 77.2% Feb 2025 Lab model card
39 Qwen3.5-2B 73.2% Mar 2026 Lab model card
40 Qwen-14B 71.7% Sep 2023 Lab model card
41 ChatGLM3-6B 69% Oct 2023 Lab model card
42 Qwen3-1.7B 68.1% May 2025 Lab model card
43 DeepSeek LLM 67B 65.2% Nov 2023 Lab model card
44 Qwen-7B 59.7% Sep 2023 Lab model card
45 Claude 1.3 54.2% Five-shot Apr 2023 Official leaderboard
46 Qwen-1_8B 50.7% Sep 2023 Lab model card
47 Qwen3.5-0.8B 50.5% Mar 2026 Lab model card
48 ChatGLM-6B 38.9% Zero-shot Mar 2023 Official leaderboard
49 LLaMA-65B 38.8% Five-shot Feb 2023 Official leaderboard

The benchmark's own leaderboard has 13 entries, 3 of them models we track. Switch the scores above to rank them together.

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
Claude 1.3
Zero-shot Five-shot
54.2%
ChatGLM-6B
Five-shot Zero-shot
38.9%
LLaMA-65B
Zero-shot Five-shot
38.8%

Source: Official C-Eval leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude 1.3 Five-shot 54.2%
Claude 1.3 Zero-shot 50.5%
ChatGLM-6B Zero-shot 38.9%
LLaMA-65B Five-shot 38.8%
LLaMA-65B Zero-shot 34.7%
ChatGLM-6B Five-shot 34.5%

5 models have held number one in labs’ own numbers since its release in May 2023 Number one over time 6 models have held number one with claims included since its release in May 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.

Anthropic Qwen DeepSeek
Own leaderboard Not tested since May 2023
Claude 1.3 54.2%
Lab claims Passed 85% in Dec 2024, by DeepSeek-V3
Qwen3-Max 93.7%

Sources: Official C-Eval leaderboard and labs’ model cards

DeepSeek’s own number for DeepSeek-V3 passed 85% in December 2024

All the scores we have, one dot each, so a model tested at several settings appears more than once.

On the benchmark’s own leaderboard Reported by the lab in its model card Best so far
0% 20% 40% 60% 80% 100% 2023 2024 2025 2026 Near the ceiling (85%) Qwen3-Max DeepSeek-R1 DeepSeek-V3 Qwen1.5-72B Qwen-14B Claude 1.3 ChatGLM-6B 0% 20% 40% 60% 80% 100% 2023 2024 2025 2026 Near the ceiling (85%) Qwen3-Max DeepSeek-V3

Sources: Official C-Eval leaderboard and labs’ model cards

Labs headline it less often now, last in September 2026

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.

Qwen
Z.ai
DeepSeek
Moonshot AI
Tencent
Baidu
Shanghai AI Lab
Meituan
Kuaishou
DateLaunch postScoreAs the post writes it
Sep 10, 2026 DeepSeek V4.1 Flash 92.1 C-Eval (EM)
May 25, 2026 Keye-VL-2.0-30B-A3B 89.2 C-Eval
Apr 24, 2026 DeepSeek-V4-Pro 93.1 C-Eval (EM)
Apr 22, 2026 Qwen3.6 27B 91.4 C-Eval
Apr 15, 2026 Qwen 3.6 35B-A3B 90 C-Eval
Apr 2, 2026 Qwen 3.6 Plus 93.3 C-Eval
Mar 30, 2026 Qwen3.5-Omni-Plus 92 C-Eval
Mar 2, 2026 Qwen3.5-0.8B 50.5 C-Eval (Thinking)
Mar 2, 2026 Qwen3.5-2B 73.2 C-Eval (Thinking)
Mar 2, 2026 Qwen3.5-4B 85.1 C-Eval
Mar 2, 2026 Qwen3.5-9B 88.2 C-Eval
Feb 24, 2026 Qwen3.5-35B-A3B 90.2 C-Eval
Feb 24, 2026 Qwen3.5-27B 90.5 C-Eval
Feb 24, 2026 Qwen3.5-122B-A10B 91.9 C-Eval
Feb 16, 2026 Qwen3.5 397B-A17B 93 C-Eval
Feb 2, 2026 Qwen3-Coder-Next 88.2 C-Eval, 88.2
Jan 23, 2026 Qwen3-Max 93.7 C-Eval
Oct 31, 2025 LongCat-Flash-Omni 91.7 CEval(acc)
Aug 29, 2025 LongCat-Flash-Chat 90.4 CEval(acc)
Aug 25, 2025 InternVL3.5-241B-A28B 90.9 C-Eval
Jun 30, 2025 ERNIE-4.5-300B-A47B 90.6 C-Eval
Apr 29, 2025 Qwen3-235B-A22B 89.6 C-Eval
Mar 21, 2025 Hunyuan-T1 91.8 CEval
Mar 16, 2025 ERNIE 4.5 C-Eval
Feb 27, 2025 Hunyuan-TurboS 90.9 C-Eval
Feb 22, 2025 Moonlight-16B-A3B 77.2 C-Eval
Jan 28, 2025 Qwen2.5-Max C-Eval
Jan 20, 2025 Kimi k1.5 88.3 C-Eval (EM), short-CoT
Jan 20, 2025 DeepSeek-R1 91.8 C-Eval (EM)
Dec 26, 2024 DeepSeek-V3 86.5 C-Eval (EM)
Jun 7, 2024 Qwen2-72B 83.8 C-Eval
May 7, 2024 DeepSeek-V2 (MoE-236B, May 2024) 78 C-Eval
Feb 5, 2024 Qwen1.5-72B 84.1 C-Eval
Nov 29, 2023 DeepSeek LLM 67B 65.2 C-Eval
Oct 27, 2023 ChatGLM3-6B 69 C-Eval
Sep 25, 2023 Qwen-14B 71.7 C-Eval (val, 5-shot)

26 system cards report it, from 9 labs

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.

DateDocumentScoreAs the document writes itSection
Sep 10, 2026 DeepSeek-V4.1-Flash Base model card 92.1 C-Eval, EM, 5-shot Capabilities
Jun 9, 2026 Keye-VL-2.0-30B-A3B technical report 89.2 C-Eval (Figure 5, printed value) Capabilities
Apr 17, 2026 Qwen3.5-Omni-Plus technical report 92 C-Eval (Table 4) Capabilities
Feb 17, 2026 GLM-5 technical report 88.8 GLM-5-Base, C-Eval (EM) (Table 11) Capabilities
Oct 31, 2025 LongCat-Flash-Omni technical report 91.7 CEval (acc) (Table 13) Capabilities
Oct 21, 2025 Ring-1T technical report 91.5 C-Eval (Table 1) Capabilities
Sep 1, 2025 LongCat-Flash-Chat technical report 90.4 CEval (acc) (Table 3) Capabilities
Aug 25, 2025 InternVL3.5-241B-A28B technical report 90.9 C-Eval (Table 2, Table 14) Capabilities
Aug 8, 2025 GLM-4.5 technical report 86.9 GLM-4.5-Base, C-Eval (EM) (Table 2) Capabilities
Jun 29, 2025 ERNIE-4.5-300B-A47B technical report 90.6 C-Eval (Table 5) Capabilities
May 14, 2025 Qwen3-8B technical report 83.4 C-Eval, thinking mode (Table 17). Non-thinking 77.9 (Table 18) Capabilities
May 14, 2025 Qwen3-4B technical report 77.5 C-Eval, thinking mode (Table 17). Non-thinking 72.2 (Table 18) Capabilities
May 14, 2025 Qwen3-32B technical report 87.3 C-Eval, thinking mode (Table 13). Non-thinking 83.3 (Table 14) Capabilities
May 14, 2025 Qwen3-30B-A3B technical report 86.6 C-Eval, thinking mode (Table 15). Non-thinking 82.9 (Table 16) Capabilities
May 14, 2025 Qwen3-235B-A22B technical report 89.6 C-Eval, thinking mode (Table 11). Non-thinking 86.1 (Table 12) Capabilities
May 14, 2025 Qwen3-14B technical report 86.2 C-Eval, thinking mode (Table 15). Non-thinking 81.0 (Table 16) Capabilities
May 14, 2025 Qwen3-1.7B technical report 68.1 C-Eval, thinking mode (Table 19). Non-thinking 61.0 (Table 20) Capabilities
Jan 20, 2025 Kimi k1.5 technical report 88.3 C-Eval (EM), short-CoT (Table 3) Capabilities
Jan 20, 2025 DeepSeek-R1 technical report 91.8 C-Eval (EM), DeepSeek-R1 (Table 8, same as final column of Table 3) Capabilities
Dec 26, 2024 DeepSeek-V3 technical report 86.5 C-Eval (EM), DeepSeek-V3 (Table 6) Capabilities
Jun 18, 2024 ChatGLM3-6B technical report 69 ChatGLM3-6B-Base, C-Eval (Table 1) Capabilities
May 7, 2024 DeepSeek-V2 (MoE-236B, May 2024) technical report 78 C-Eval 5-shot, DeepSeek-V2 Chat (RL) (Table 3) Capabilities
Jan 5, 2024 DeepSeek LLM 67B technical report 65.2 C-Eval 0-shot, DeepSeek 67B Chat (Table 6) Capabilities
Sep 28, 2023 Qwen-7B technical report 59.7 C-Eval 0-shot, higher of the two settings, 5-shot 59.3 (Table 5) Capabilities
Sep 28, 2023 Qwen-14B technical report 71.7 C-Eval 5-shot, higher of the two settings, 0-shot 69.8 (Table 5) Capabilities
Sep 28, 2023 Qwen-1_8B technical report 50.7 C-Eval 0-shot, higher of the two settings, 5-shot 50.3 (Table 5) Capabilities

More in Knowledge

All 121 in Knowledge
  • MMLU-Pro In use Harder 10-option version of MMLU with reasoning-focused questions across 14 subjects. Intern-S2-397B 89.8%
  • 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%
  • KMMLU-Pro In use Models answer multiple-choice questions taken from Korean National Professional Licensure exams. It tests professional knowledge used in Korean industry. Solar Pro 4 79.2%
  • HLE-Verified In use Verified and revised version of Humanity's Last Exam (1,811 items) with uncertain items removed. Seed2.0 Pro 73.6%