Compass Academic Leaderboard
Models answer academic reasoning tasks. OpenCompass publishes model rankings and predictions across the underlying evaluations.
The leaderboard aggregates results from component evaluations.
- Made by
- OpenCompass
- Released
- December 2024
- Kind
- Composite indices
- Level
- University
- Scored
- Settled by events
- Answers
- Short answer
- Tasks
- All public
- Measures
- Capability
- Interaction
- Single turn
- Items from
- Another benchmark
- Refresh
- One fixed set
- Run by
- Anyone
- Strongest model tested
-
DeepSeek-V3 #155 on the capability index
- Last model added
- March 2025
- Scores refreshed
- November 20th, 2026
DeepSeek-V3 leads with 72.6% among models released by March 2025 DeepSeek-V3 leads with 72.6% 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:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 72.6
DeepSeek-V3
- 71.8
Qwen2.5-Max
- 68.5
Qwen2.5-72B
- 68.0
Qwen2.5-32B
- 67.6
DeepSeek-V2.5-1210
- 67.2
Llama 3.3 70B Instruct
- 67.1
MiniMax-Text-01
- 64.5
DeepSeek-V2.5
- 64.3
Qwen2.5-14B
- 64.3
Gemma 3 27B
- 56.5
Qwen2.5-7B
- 56.0
Gemma 2 27B
- 51.0 InternLM2.5-Chat-7B
- 50.7
Llama 3.1-8B
- 50.3
Gemma 2 9B
- 42.9
Llama 3.2 3B
- 1
DeepSeek-V3 72.6%
- 2
Qwen2.5-Max 71.8%
- 3
Qwen2.5-72B 68.5%
- 4
Qwen2.5-32B 68%
- 5
DeepSeek-V2.5-1210 67.6%
- 6
Llama 3.3 70B Instruct 67.2%
- 7
MiniMax-Text-01 67.1%
- 8
DeepSeek-V2.5 Sep 2024 64.5%
- 9
Qwen2.5-14B 64.3%
- 10
Gemma 3 27B 64.3%
- 11
Qwen2.5-7B 56.5%
- 12
Gemma 2 27B 56%
- 13 InternLM2.5-Chat-7B 51%
- 14
Llama 3.1-8B 50.7%
- 15
Gemma 2 9B 50.3%
- 16
Llama 3.2 3B 42.9%
Source: Official Compass Academic Leaderboard leaderboard
- 1
DeepSeek-V3 72.6%
- 2
Qwen2.5-Max 71.8%
- 3
Qwen2.5-72B 68.5%
- 4
Qwen2.5-32B 68%
- 5
DeepSeek-V2.5-1210 67.6%
- 6
Llama 3.3 70B Instruct 67.2%
- 7
MiniMax-Text-01 67.1%
- 8
DeepSeek-V2.5 Sep 2024 64.5%
- 9
Qwen2.5-14B 64.3%
- 10
Gemma 3 27B 64.3%
- 11
Qwen2.5-7B 56.5%
- 12
Gemma 2 27B 56%
- 13
Llama 3.1-8B 50.7%
- 14
Gemma 2 9B 50.3%
- 15
Llama 3.2 3B 42.9%
Source: Official Compass Academic Leaderboard leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 72.6% | Dec 2024 | Official leaderboard | |
| 2 | | 71.8% | Jan 2025 | Official leaderboard | |
| 3 | | 68.5% | Sep 2024 | Official leaderboard | |
| 4 | | 68% | Sep 2024 | Official leaderboard | |
| 5 | | 67.6% | Dec 2024 | Official leaderboard | |
| 6 | | 67.2% | Dec 2024 | Official leaderboard | |
| 7 | | 67.1% | Jan 2025 | Official leaderboard | |
| 8 | | 64.5% | Sep 2024 | Official leaderboard | |
| 9 | | 64.3% | Sep 2024 | Official leaderboard | |
| 10 | | 64.3% | Mar 2025 | Official leaderboard | |
| 11 | | 56.5% | Sep 2024 | Official leaderboard | |
| 12 | | 56% | Jun 2024 | Official leaderboard | |
| 13 | InternLM2.5-Chat-7B open weights, 7B | 51% | Official leaderboard | ||
| 14 | | 50.7% | Jul 2024 | Official leaderboard | |
| 15 | | 50.3% | Jun 2024 | Official leaderboard | |
| 16 | | 42.9% | Sep 2024 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 72.6% | Dec 2024 | Official leaderboard | |
| 2 | | 71.8% | Jan 2025 | Official leaderboard | |
| 3 | | 68.5% | Sep 2024 | Official leaderboard | |
| 4 | | 68% | Sep 2024 | Official leaderboard | |
| 5 | | 67.6% | Dec 2024 | Official leaderboard | |
| 6 | | 67.2% | Dec 2024 | Official leaderboard | |
| 7 | | 67.1% | Jan 2025 | Official leaderboard | |
| 8 | | 64.5% | Sep 2024 | Official leaderboard | |
| 9 | | 64.3% | Sep 2024 | Official leaderboard | |
| 10 | | 64.3% | Mar 2025 | Official leaderboard | |
| 11 | | 56.5% | Sep 2024 | Official leaderboard | |
| 12 | | 56% | Jun 2024 | Official leaderboard | |
| 13 | | 50.7% | Jul 2024 | Official leaderboard | |
| 14 | | 50.3% | Jun 2024 | Official leaderboard | |
| 15 | | 42.9% | Sep 2024 | Official leaderboard |
Rankings by part
We read 15 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as IFEval, BBH and GPQA_diamond. Qwen2.5-Max leads 6 of them, and 4 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
IFEval
15 models we track have a score here.
- 1
Llama 3.3 70B Instruct ↑5 88.5%
- 2
DeepSeek-V3 ↓1 85.8%
- 3
Qwen2.5-Max ↓1 84.7%
- 4
MiniMax-Text-01 ↑3 83.5%
- 5
Qwen2.5-72B ↓2 83%
- 6
Gemma 3 27B ↑4 81%
- 7
DeepSeek-V2.5-1210 ↓2 79.3%
- 8
Qwen2.5-32B ↓4 78.9%
- 9
Gemma 2 27B ↑3 78.2%
- 10
DeepSeek-V2.5 Sep 2024 ↓2 77.6%
BBH
15 models we track have a score here.
- 1
Qwen2.5-Max ↑1 89.6%
- 2
DeepSeek-V2.5-1210 ↑3 85.1%
- 3
Llama 3.3 70B Instruct ↑3 84.2%
- 4
Qwen2.5-72B ↓1 82.5%
- 5
DeepSeek-V3 ↓4 81.5%
- 6
Qwen2.5-32B ↓2 81%
- 7
MiniMax-Text-01 80.9%
- 8
DeepSeek-V2.5 Sep 2024 80.7%
- 9
Gemma 3 27B ↑1 77.9%
- 10
Qwen2.5-14B ↓1 74.9%
GPQA_diamond
15 models we track have a score here.
- 1
MiniMax-Text-01 ↑6 58.1%
- 2
DeepSeek-V3 ↓1 57.1%
- 3
Qwen2.5-Max ↓1 54%
- 4
Qwen2.5-72B ↓1 52%
- 5
Gemma 3 27B ↑5 51%
- 6
Llama 3.3 70B Instruct 50%
- 7
Qwen2.5-32B ↓3 49.5%
- 8
DeepSeek-V2.5-1210 ↓3 47%
- 9
DeepSeek-V2.5 Sep 2024 ↓1 42.4%
- 10
Qwen2.5-14B ↓1 41.9%
Math-500
15 models we track have a score here.
- 1
DeepSeek-V3 89.6%
- 2
Gemma 3 27B ↑8 86.4%
- 3
Qwen2.5-72B 84.8%
- 4
DeepSeek-V2.5-1210 ↑1 84.6%
- 5
Qwen2.5-32B ↓1 82.4%
- 6
Qwen2.5-14B ↑3 81%
- 7
Qwen2.5-Max ↓5 79.6%
- 8
MiniMax-Text-01 ↓1 76.8%
- 9
Qwen2.5-7B ↑2 76.6%
- 10
Llama 3.3 70B Instruct ↓4 76%
AIME2024
15 models we track have a score here.
- 1
DeepSeek-V3 26.7%
- 2
Qwen2.5-32B ↑2 23.3%
- 3
Qwen2.5-72B 23.3%
- 4
Gemma 3 27B ↑6 23.3%
- 5
Qwen2.5-14B ↑4 20%
- 6
Llama 3.3 70B Instruct 20%
- 7
Qwen2.5-Max ↓5 20%
- 8
DeepSeek-V2.5-1210 ↓3 16.7%
- 9
DeepSeek-V2.5 Sep 2024 ↓1 13.3%
- 10
MiniMax-Text-01 ↓3 10%
MMLU-Pro
15 models we track have a score here.
- 1
DeepSeek-V3 75.9%
- 2
MiniMax-Text-01 ↑5 75.1%
- 3
Qwen2.5-Max ↓1 73.6%
- 4
Qwen2.5-72B ↓1 71.3%
- 5
Llama 3.3 70B Instruct ↑1 71.3%
- 6
Qwen2.5-32B ↓2 68.6%
- 7
Gemma 3 27B ↑3 67.8%
- 8
DeepSeek-V2.5-1210 ↓3 67.2%
- 9
DeepSeek-V2.5 Sep 2024 ↓1 65.2%
- 10
Qwen2.5-14B ↓1 64.5%
LiveCodeBench
15 models we track have a score here.
- 1
Qwen2.5-Max ↑1 66.5%
- 2
DeepSeek-V3 ↓1 65.5%
- 3
DeepSeek-V2.5-1210 ↑2 58.3%
- 4
Qwen2.5-72B ↓1 57.3%
- 5
Qwen2.5-32B ↓1 53.8%
- 6
DeepSeek-V2.5 Sep 2024 ↑2 52%
- 7
Qwen2.5-14B ↑2 48%
- 8
Llama 3.3 70B Instruct ↓2 40%
- 9
Qwen2.5-7B ↑2 38.3%
- 10
MiniMax-Text-01 ↓3 36.3%
HumanEval
15 models we track have a score here.
- 1
Qwen2.5-Max ↑1 92.7%
- 2
DeepSeek-V3 ↓1 92.1%
- 3
MiniMax-Text-01 ↑4 89%
- 4
Qwen2.5-32B 88.4%
- 5
DeepSeek-V2.5-1210 87.8%
- 6
DeepSeek-V2.5 Sep 2024 ↑2 87.2%
- 7
Llama 3.3 70B Instruct ↓1 87.2%
- 8
Gemma 3 27B ↑2 87.2%
- 9
Qwen2.5-7B ↑2 84.8%
- 10
Qwen2.5-72B ↓7 84.2%
Drop
15 models we track have a score here.
- 1
DeepSeek-V3 92.1%
- 2
Llama 3.3 70B Instruct ↑4 90.2%
- 3
Gemma 3 27B ↑7 89.7%
- 4
MiniMax-Text-01 ↑3 89.4%
- 5
DeepSeek-V2.5-1210 89.1%
- 6
Qwen2.5-Max ↓4 88.8%
- 7
Qwen2.5-32B ↓3 88%
- 8
Qwen2.5-72B ↓5 87.5%
- 9
Qwen2.5-14B 85.7%
- 10
Gemma 2 27B ↑2 83.3%
Hellaswag
15 models we track have a score here.
- 1
Qwen2.5-Max ↑1 93.6%
- 2
Qwen2.5-32B ↑2 92.1%
- 3
DeepSeek-V2.5-1210 ↑2 91.2%
- 4
DeepSeek-V2.5 Sep 2024 ↑4 90.7%
- 5
Qwen2.5-14B ↑4 90.7%
- 6
Llama 3.3 70B Instruct 89%
- 7
MiniMax-Text-01 88.5%
- 8
Qwen2.5-72B ↓5 87.2%
- 9
Gemma 2 27B ↑3 86.8%
- 10
Gemma 2 9B ↑4 85.3%
MUSR
15 models we track have a score here.
- 1
DeepSeek-V2.5 Sep 2024 ↑7 74.5%
- 2
DeepSeek-V3 ↓1 74%
- 3
Qwen2.5-Max ↓1 73.4%
- 4
Llama 3.3 70B Instruct ↑2 73.1%
- 5
Qwen2.5-32B ↓1 69.4%
- 6
Qwen2.5-72B ↓3 69.3%
- 7
Gemma 3 27B ↑3 69.2%
- 8
MiniMax-Text-01 ↓1 68.5%
- 9
Qwen2.5-14B 65%
- 10
DeepSeek-V2.5-1210 ↓5 62.8%
KorBench
15 models we track have a score here.
- 1
DeepSeek-V3 63.4%
- 2
DeepSeek-V2.5-1210 ↑3 58.9%
- 3
Qwen2.5-Max ↓1 55.5%
- 4
MiniMax-Text-01 ↑3 55.1%
- 5
Qwen2.5-32B ↓1 54.8%
- 6
Llama 3.3 70B Instruct 52.6%
- 7
Gemma 3 27B ↑3 51.8%
- 8
Qwen2.5-72B ↓5 51.4%
- 9
Qwen2.5-14B 49.8%
- 10
DeepSeek-V2.5 Sep 2024 ↓2 46.9%
CMMLU
15 models we track have a score here.
- 1
Qwen2.5-Max ↑1 87.1%
- 2
DeepSeek-V3 ↓1 85.2%
- 3
Qwen2.5-72B 85%
- 4
MiniMax-Text-01 ↑3 83.9%
- 5
Qwen2.5-32B ↓1 82.9%
- 6
DeepSeek-V2.5-1210 ↓1 80.2%
- 7
Qwen2.5-14B ↑2 80.1%
- 8
DeepSeek-V2.5 Sep 2024 78.6%
- 9
Qwen2.5-7B ↑2 77.3%
- 10
Llama 3.3 70B Instruct ↓4 73%
MMLU
15 models we track have a score here.
- 1
DeepSeek-V3 89%
- 2
MiniMax-Text-01 ↑5 88.9%
- 3
Qwen2.5-Max ↓1 87.5%
- 4
Llama 3.3 70B Instruct ↑2 86.4%
- 5
Qwen2.5-72B ↓2 86.3%
- 6
Qwen2.5-32B ↓2 84%
- 7
DeepSeek-V2.5-1210 ↓2 83.3%
- 8
Gemma 3 27B ↑2 82.4%
- 9
DeepSeek-V2.5 Sep 2024 ↓1 82.4%
- 10
Qwen2.5-14B ↓1 81.5%
BigCodeBench
15 models we track have a score here.
- 1
Qwen2.5-Max ↑1 29.7%
- 2
DeepSeek-V3 ↓1 26.4%
- 3
Llama 3.3 70B Instruct ↑3 25.7%
- 4
Qwen2.5-32B 23.6%
- 5
MiniMax-Text-01 ↑2 23%
- 6
Gemma 3 27B ↑4 23%
- 7
Qwen2.5-72B ↓4 22.3%
- 8
DeepSeek-V2.5-1210 ↓3 22.3%
- 9
Llama 3.1-8B ↑4 20.3%
- 10
DeepSeek-V2.5 Sep 2024 ↓2 20.3%
2 models have held number one on the official leaderboard since its release in December 2024
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official Compass Academic Leaderboard leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official Compass Academic 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 Other
- Artificial Analysis Intelligence Index v5 In use Version 5 of Artificial Analysis's index: nine evaluations in four equal categories, most with private test sets, run three times each in its own agent harness. Claude Fable 5.2 64.6
- Personal Superintelligence Index (Meta) In use Meta's composite of seven personal and knowledge-work evals: Agents' Last Exam, AutomationBench-AA, FocusBench, JobBench, GDP.pdf, VBT and DRACO.
- Vals Science Index In use Agents are scored on the equal average of BioMysteryBench, MysteryMechanism and Terminal-Bench-Science. GPT-6.2 Astra 68.4%
- Decision Index Public v0.3 In use Decision models return probabilities over labeled options in public tasks covering knowledge, language, retrieval, tools, and arts. A chance-corrected weighted aggregate combines the task scores. Granite Switch 4.1 30B 51.3
- SAGE v1.1 (Vals) In use Models grade handwritten undergraduate math and computer science work against rubrics. Scored by class-balanced accuracy of each rubric check against expert grading. Claude Opus 5.5 91.3%
- MTEB Fading Massive Text Embedding Benchmark: retrieval, clustering, classification across languages and modalities.
Qwen3-Embedding-8B 72.2%
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 27B. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
- Pass Independent runs Scores from the official leaderboard (16 entries).
- Pass Models scored 15 models we track, plus 1 other entry on its leaderboard.
- Warning Newest score The newest came 20 months ago, in March 2025, for Gemma 3 27B.
- Note Frontier models None of the top 10 has a score. The strongest tested is DeepSeek-V3 (#155).
- Pass Headroom The best is 72.6%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on huggingface.co: 15 rows for models we track and 1 other entry 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.