CommonSenseQA
Models answer five-way multiple-choice questions built from ConceptNet relations that need everyday background knowledge. Scored by accuracy.
Test labels are hidden, so model cards usually report the dev set.
In the authors’ words
When answering a question, people often draw upon their rich world knowledge in addition to the particular context. Recent work has focused primarily on answering questions given some relevant document or context, and required very little general background. To investigate question answering with prior knowledge, we present CommonsenseQA: a challenging new dataset for commonsense question answering. To capture common sense beyond associations, we extract from ConceptNet (Speer et al., 2017) multiple target concepts that have the same semantic relation to a single source concept. Crowd-workers are asked to author multiple-choice questions that mention the source concept and discriminate in turn between each of the target concepts. This encourages workers to create questions with complex semantics that often require prior knowledge. We create 12,247 questions through this procedure and demonstrate the difficulty of our task with a large number of strong baselines. Our best baseline is based on BERT-large (Devlin et al., 2018) and obtains 56% accuracy, well below human performance, which is 89%.
- Level
- Everyday Most adults, without training
- Scored
- Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
- Answers
- Multiple choice
PaLM 2-L leads with 90.4% among models released by Apr 2025
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.
The newest model with a score here came out in Apr 2025, and nobody has reported a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6 Astra
- Claude Opus 5.5
- GPT-6.1 Terra
- GPT-6.1 Sol
- and 16 more
- 90.4 PaLM 2-L
- 86.5 ChatGLM3-6B
- 85.8
Llama 3.1-405B - 84.1 Llama 3.1-70B
- 83.8 Llama 3-70B
- 82.8
phi-3-medium 14B - 80.2 phi-3-mini 3.8B
- 80.0 phi-3-small 7.4B
- 78.5 Llama 2-70B
- 75.0 Llama 3.1-8B
- 74.3 Llama 2-34B
- 72.6 Llama 3-8B
- 71.6 BitNet b1.58 2B4T
- 71.3
Gemma 7B - 67.3 Llama 2-13B
- 65.3 Gemma 2B
- 57.8 Llama 2-7B
- 1 PaLM 2-L 90.4%
- 2 ChatGLM3-6B 86.5%
- 3 Llama 3.1-405B 85.8%
- 4 Llama 3.1-70B 84.1%
- 5 Llama 3-70B 83.8%
- 6 phi-3-medium 14B 82.8%
- 7 phi-3-mini 3.8B 80.2%
- 8 phi-3-small 7.4B 80%
- 9 Llama 2-70B 78.5%
- 10 Llama 3.1-8B 75%
- 11 Llama 2-34B 74.3%
- 12 Llama 3-8B 72.6%
- 13 BitNet b1.58 2B4T 71.6%
- 14 Gemma 7B 71.3%
- 15 Llama 2-13B 67.3%
- 16 Gemma 2B 65.3%
- 17 Llama 2-7B 57.8%
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | PaLM 2-L | 90.4% | May 2023 | Model card | |
| 2 | ChatGLM3-6B | 86.5% | Jun 2024 | Model card | |
| 3 | Llama 3.1-405B | 85.8% | Jul 2024 | Model card | |
| 4 | Llama 3.1-70B | 84.1% | Jul 2024 | Model card | |
| 5 | Llama 3-70B | 83.8% | Apr 2024 | Model card | |
| 6 | phi-3-medium 14B | 82.8% | Aug 2024 | Model card | |
| 7 | phi-3-mini 3.8B | 80.2% | Aug 2024 | Model card | |
| 8 | phi-3-small 7.4B | 80% | Aug 2024 | Model card | |
| 9 | Llama 2-70B | 78.5% | Jul 2023 | Model card | |
| 10 | Llama 3.1-8B | 75% | Jul 2024 | Model card | |
| 11 | Llama 2-34B | 74.3% | Jul 2023 | Model card | |
| 12 | Llama 3-8B | 72.6% | Apr 2024 | Model card | |
| 13 | BitNet b1.58 2B4T | 71.6% | Apr 2025 | Model card | |
| 14 | Gemma 7B | 71.3% | Feb 2024 | Model card | |
| 15 | Llama 2-13B | 67.3% | Jul 2023 | Model card | |
| 16 | Gemma 2B | 65.3% | Mar 2024 | Model card | |
| 17 | Llama 2-7B | 57.8% | Jul 2023 | Model card |
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
The best score passed 85% in May 2023
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Labs dropped it from launch posts once the best score passed 85%
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.
| Date | Launch post | Score | As the post writes it |
|---|---|---|---|
| 15 Apr 2025 | BitNet b1.58 2B4T | 71.6 | CommonsenseQA |
| 22 Apr 2024 | phi-3-medium 14B | 82.6 | CommonSenseQA (10-Shot) |
| 18 Apr 2024 | Llama 3-70B | 83.8 | CommonSenseQA (7-shot) [pretrained] |
| 21 Feb 2024 | Gemma 7B | 71.3 | CommonsenseQA (7-shot) |
| 10 May 2023 | PaLM 2-L | 90.4 | CSQA (7), CoT + SC |
19 system cards report it, from 4 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.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| 16 Apr 2025 | BitNet b1.58 2B4T technical report | 71.6 | CommonsenseQA (10-shot; Acc) (Table 1, Table 3) | Capabilities |
| 23 Nov 2024 | Llama 3.1-8B technical report | 75 | CommonSenseQA, pretrained (Table 11) | Capabilities |
| 23 Nov 2024 | Llama 3.1-70B technical report | 84.1 | CommonSenseQA, pretrained (Table 11) | Capabilities |
| 23 Nov 2024 | Llama 3.1-405B technical report | 85.8 | CommonSenseQA, pretrained (Table 11) | Capabilities |
| 30 Aug 2024 | phi-3-small 7.4B technical report | 80 | CommonSenseQA (10-Shot) (Section 3 table) | Capabilities |
| 30 Aug 2024 | phi-3-mini 3.8B technical report | 80.2 | CommonSenseQA (10-Shot) (Section 3 table) | Capabilities |
| 30 Aug 2024 | phi-3-medium 14B technical report | 82.8 | CommonSenseQA (10-Shot) (Section 3 table) | Capabilities |
| 23 Jul 2024 | Llama 3.1-8B model card | 75 | CommonSenseQA, 7-shot, acc_char, pretrained (Base pretrained models table) | Capabilities |
| 23 Jul 2024 | Llama 3.1-70B model card | 84.1 | CommonSenseQA, 7-shot, acc_char, pretrained (Base pretrained models table) | Capabilities |
| 23 Jul 2024 | Llama 3.1-405B model card | 85.8 | CommonSenseQA, 7-shot, acc_char, pretrained (Base pretrained models table) | Capabilities |
| 18 Jun 2024 | ChatGLM3-6B technical report | 86.5 | ChatGLM3-6B-Base, CommonSenseQA (Table 1) | Capabilities |
| 18 Apr 2024 | Llama 3-8B model card | 72.6 | CommonSenseQA (7-shot), pretrained (Base pretrained models table) | Capabilities |
| 18 Apr 2024 | Llama 3-70B model card | 83.8 | CommonSenseQA (7-shot), pretrained (Base pretrained models table) | Capabilities |
| 13 Mar 2024 | Gemma 7B technical report | 71.3 | CQA, PT, 7-shot (Table 6) | Capabilities |
| 13 Mar 2024 | Gemma 2B technical report | 65.3 | CQA, PT, 7-shot (Table 6) | Capabilities |
| 19 Jul 2023 | Llama 2-7B technical report | 57.8 | CSQA, 7-shot, pretrained (Table 20) | Capabilities |
| 19 Jul 2023 | Llama 2-70B technical report | 78.5 | CSQA, 7-shot, pretrained (Table 20) | Capabilities |
| 19 Jul 2023 | Llama 2-34B technical report | 74.3 | CSQA, 7-shot, pretrained (Table 20) | Capabilities |
| 19 Jul 2023 | Llama 2-13B technical report | 67.3 | CSQA, 7-shot, pretrained (Table 20) | Capabilities |
Versions
We switch the top row to each new version when it comes out. Open it for one row per version. An old version stays in play while people still test it, so whoever first takes it past 85% gets credit for closing it out.
More 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% - 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% - BIG-Bench Extra Hard In use Models solve harder replacements for each of the 23 BIG-Bench Hard tasks, 4,520 examples in total. The score is a harmonic mean of task accuracies. Gemma 4 31B IT 74.4%
- 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%