C-Eval: leaderboard
C-Eval: Evaluates broad language-model knowledge, reasoning, commonsense, instruction following, or exam-style accuracy.
Metric: Average (%). Source: github.com. Source: benchmarklist.com. Saturation forecast: Estimated already saturated. 143 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Yi 1.5 34B | 83.1 | #807 |
| 2 | Qwen-72B | 82.8 | #779 |
| 3 | Yi 34B (Base) | 81.4 | #972 |
| 4 | OrionStar-Yi-34B-Chat | 78.1 | #890 |
| 5 | Qwen-14B | 72.1 | #1171 |
| 6 | Yi 6B (Base) | 72 | #1472 |
| 7 | GPT-4 | 68.7 | #438 |
| 8 | Qwen-7B | 59.6 | #1402 |
| 9 | Qwen-1.8B | 54.7 | #1546 |
| 10 | ChatGPT | 54.4 | #900 |
| 11 | Claude v1.3 | 54.2 | #1001 |
| 12 | internlm-7B | 52.8 | #1517 |
| 13 | ChatGLM2 6B | 51.7 | #1498 |
| 14 | GLM-130B | 44 | #1568 |
| 15 | Baichuan-7B | 42.8 | #1527 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=c-eval · How It Works · Data refreshed daily, snapshot 2026-10-20.