Qwen-7B: benchmark results
Alibaba Cloud's original Qwen-7B base model (Tongyi Qianwen), pretrained on 2.4 trillion tokens of web text, books and code with a roughly 150,000-token vocabulary. Provider: Alibaba. Released 2023-08-03. Access: Open.
Aggregate Rating 1341 ± 17, rank #1402 of 1656 rated models, from 40 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| ARC Challenge (AI2) | 75.3 | Accuracy (%) | 72.4 |
| T-Eval | 59.5 | Overall Score (%) | 70 |
| CFLUE | 43.63 | Accuracy | 62.5 |
| Open LLM Leaderboard v1 - GSM8K | 44.96 | Accuracy (%) (5-shot) | 59 |
| GSM8K | 51.7 | Accuracy (%) | 52.7 |
| Open LLM Leaderboard v1 - MMLU | 59.84 | Accuracy (%) (5-shot) | 46.1 |
| CyberMetric | 58 | 500-question accuracy (%) | 41.7 |
| CMMLU | 58.66 | 5-shot Avg Accuracy (%) | 40 |
| Big-Bench Hard | 45 | Average (%) | 39.6 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 47.79 | MC2 (%) (0-shot) | 38.7 |
| BoolQ | 76.4 | Accuracy (%) | 38.3 |
| VMLU | 32.81 | Average (%) | 37.5 |
Interactive version: theaggregate.ai/model?slug=qwen-7b · How It Works · Data refreshed daily, snapshot 2026-10-20.