T-Eval — leaderboard
Metric: Overall Score (%). Source: open-compass.github.io. 21 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4 Turbo (Preview) | 86.4 |
| 2 | GPT-3.5 Turbo (1106) | 84 |
| 3 | Claude 2.1 | 78.8 |
| 4 | Qwen-14B | 66.3 |
| 5 | Qwen-7B | 59.5 |
| 6 | Mistral 7B | 56 |
| 7 | Llama 2 70B | 53 |
| 8 | chatglm3-6B | 51.4 |
| 9 | vicuna-13B | 48.1 |
| 10 | internlm-7B | 45.8 |
| 11 | Vicuna-7B | 44.8 |
| 12 | Llama 2 13B | 37.3 |
| 13 | Llama 2 7B | 27.4 |
Interactive version: theaggregate.ai/benchmark?slug=t-eval · How the rankings work · Data refreshed daily, snapshot 2026-07-22.