Free🇺🇸
OpenAI's set of 8,500 grade school math word problems with worked solutions, used to test multi-step reasoning of language models.
C-Eval overview and key facts (October 2026)
Chinese evaluation suite with 13,948 multiple-choice questions in 52 disciplines and four difficulty levels for foundation models.
Free
What is C-Eval?
C-Eval is an evaluation dataset of Chinese multiple-choice exam questions. Companies serving Chinese-speaking customers use it to compare how well language models handle Chinese knowledge and reasoning.
C-Eval contains 13,948 questions in 52 disciplines, from computer networks to law and medicine, at four difficulty levels. Each subject has a dev split with five explained examples, a validation split and a test split. In July 2025 the maintainers released the complete test set, so teams can evaluate locally. The data is licensed CC BY-NC-SA 4.0, and the paper was published on arXiv in 2023.
C-Eval AI features and use cases
- Compares models for Chinese marketsTeams choosing a model for customer service or documents in Chinese can run C-Eval on candidates and see which handles Chinese content best. English benchmarks do not show this. The results complement tests with your own Chinese documents.
- Checks subject-specific strengthScores per discipline show whether a model is weaker in law, medicine or engineering. Teams can plan extra testing for the subjects that matter to their use case. This keeps the test effort focused on what matters.
- Runs few-shot evaluationsThe five explained examples per subject support standardized few-shot tests, making results comparable across models and runs. Teams can repeat the same setup after model updates and see whether performance in Chinese has changed.
What sets C-Eval apart
- Covers 52 disciplines at four difficulty levels in Chinese.
- Complete test set released in July 2025 for local evaluation.
Where C-Eval is strongest
Questions to ask before you buy C-Eval
The CC BY-NC-SA 4.0 license excludes commercial use; check with legal whether internal model selection in a company counts as commercial before using it. Results only describe Chinese-language performance. Public questions may have leaked into training data. For German or English use cases, other benchmarks fit better.
C-Eval reviews
No reviews yet. Be the first to share how your team uses it.
Every review shows its verification level. We read each one before it goes live. How we check reviews
Questions about C-Eval
May companies use C-Eval?
The dataset card lists CC BY-NC-SA 4.0, a non-commercial license. Clarify with legal counsel whether your intended use is allowed.
How is the data structured?
Each of the 52 subjects has dev, validation and test splits. The dev split holds five examples with explanations for few-shot evaluation.
Is the test set public?
Yes. Since July 27, 2025 the complete test set is available, so evaluations can run locally.
C-Eval update historyno updates yet
No changes recorded for C-Eval yet.
Changes the vendor made and confirmed user reports, each checked against the vendor's own pages. We review every tool monthly.
Alternatives to C-Eval
Tools in the same category (Evaluation data) with the most overlapping markets.
Free🇺🇸
Collaborative benchmark of 162 English legal reasoning tasks from 40 contributors, covering contracts, statutes and court opinions.
Free🇨🇦
Evaluation set of about 12,000 expert-reviewed multiple-choice questions with ten options each, used to compare language models.
Free🇺🇸
Benchmark that tests whether AI systems resolve real GitHub issues, with Verified, Lite, Multimodal, Multilingual and Bash Only boards.