reasoning benchmark
C-Eval Leaderboard
C-Eval is a Chinese evaluation suite with 13,948 multiple-choice questions across 52 disciplines and four difficulty levels, including the C-Eval Hard subset.
Updated Sep 5, 2026
Models18
Model coverage18
MetricScore
EvidenceB
C-Eval Ranking
Higher score ranks better on this benchmark.
Sort by
Rank | Model | Score | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelAC | Score93.30% | Percentile100.00% | Participants18 | EvidenceC | Evaluated |
| Rank02 | ModelAC | Score93.00% | Percentile94.12% | Participants18 | EvidenceC | Evaluated |
| Rank03 | ModelMA | Score92.50% | Percentile88.24% | Participants18 | EvidenceC | Evaluated |
| Rank04 | ModelAC | Score91.90% | Percentile82.35% | Participants18 | EvidenceC | Evaluated |
| Rank05 | ModelXI | Score91.50% | Percentile76.47% | Participants18 | EvidenceC | Evaluated |
| Rank06 | ModelAC | Score91.40% | Percentile70.59% | Participants18 | EvidenceC | Evaluated |
| Rank07 | ModelAC | Score90.50% | Percentile64.71% | Participants18 | EvidenceC | Evaluated |
| Rank08 | ModelAC | Score90.20% | Percentile58.82% | Participants18 | EvidenceC | Evaluated |
| Rank09 | ModelAC | Score90.00% | Percentile52.94% | Participants18 | EvidenceC | Evaluated |
| Rank10 | ModelMA | Score88.30% | Percentile47.06% | Participants18 | EvidenceC | Evaluated |
| Rank11 | ModelAC | Score88.20% | Percentile41.18% | Participants18 | EvidenceC | Evaluated |
| Rank12 | ModelDE | Score86.50% | Percentile35.29% | Participants18 | EvidenceC | Evaluated |
| Rank13 | ModelAC | Score85.10% | Percentile29.41% | Participants18 | EvidenceC | Evaluated |
| Rank14 | ModelAC | Score83.80% | Percentile23.53% | Participants18 | EvidenceC | Evaluated |
| Rank15 | ModelAC | Score77.20% | Percentile17.65% | Participants18 | EvidenceC | Evaluated |
| Rank16 | ModelAC | Score73.20% | Percentile11.76% | Participants18 | EvidenceC | Evaluated |
| Rank17 | ModelAC | Score50.50% | Percentile5.88% | Participants18 | EvidenceC | Evaluated |
| Rank18 | ModelBA | Score40.70% | Percentile0.00% | Participants18 | EvidenceC | Evaluated |
C-Eval Highlights
The leading models and scores on this benchmark.
C-Eval Score Distribution
A closer view of the leading scores on this benchmark.
The Top AI Models for C-Eval
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
What is C-Eval?
What C-Eval measures and how its scores work.
C-Eval is a benchmark for assessing foundation models' advanced knowledge and reasoning abilities in a Chinese context.
It measures Score, reported as a ratio, across questions covering humanities, science, and engineering at middle school, high school, college, and professional levels.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- C-Eval
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- overall
LLM Stats. Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about C-Eval.
Which model scores highest on C-Eval?
Qwen3.6 Plus is currently ranked first with 93.30%.
What are the top three models on C-Eval?
The current leaders are Qwen3.6 Plus (93.30%), Qwen3.5-397B-A17B (93.00%), and Kimi K2 Base (92.50%).
Which C-Eval model has the lowest official input price?
Qwen3.6-35B-A3B has the lowest matched official input price at $0.25 input / $1.5 output per 1M tokens.
Which models are fastest among C-Eval results?
The fastest matched records are DeepSeek-V3 (100.00 tok/s via DeepSeek) and MiMo-V2.5-Pro (5.27 tok/s via DeepInfra).
Does the highest score result prove overall model quality?
No. This benchmark measures one defined capability or task. The overall LLMBoard score uses a separate aggregation across eligible benchmark evidence.