Qwen 2 7B Instruct: benchmark results
Alibaba's Apache-2.0 7B Qwen2 instruct model (June 2024) handling context lengths up to 128K tokens. Provider: Alibaba. Released 2024-06-06. Access: Open.
Aggregate Rating 1452 ± 7, rank #991 of 1656 rated models, from 510 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| SeaEval - Multilingual Reasoning - ZBench (Zero-Shot) | 72.73 | Accuracy (%) | 100 |
| SeaEval - Cultural Reasoning - SG-Eval v2 Open (Zero-Shot) | 56.56 | Accuracy (%) | 95.7 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 61.26 | Exact match, flexible extract (%) | 94.8 |
| OpenCompass Language - NLP - Chinese (CompassBench 2407) | 59.4 | Score (%) | 94.1 |
| Open Portuguese LLM - ASSIN2 RTE | 93.91 | Macro F1 (%) | 93.5 |
| Open Korean LLM - KorNAT-Harmless | 68.1 | Normalized accuracy (%) | 91.6 |
| Open Korean LLM - KorNAT-Helpful | 51.63 | Normalized accuracy (%) | 91.4 |
| SeaEval - Fundamental NLP Tasks - MRPC (Zero-Shot) | 78.68 | Accuracy (%) | 91.3 |
| Open Korean LLM - Ko-IFEval | 45.82 | Strict accuracy, prompt/instruction mean (%) | 90.4 |
| SeaEval - Cultural Reasoning - CN-Eval (Zero-Shot) | 82.86 | Accuracy (%) | 87.5 |
| Open Arabic LLM - Arabic MMLU Computer Science (Middle School) | 92.59 | Accuracy (%) | 87 |
| Open PL LLM - PSC (generative, 5-shot) | 97.43 | Binary F1 (%) | 85.3 |
Interactive version: theaggregate.ai/model?slug=qwen-2-7b-instruct · How It Works · Data refreshed daily, snapshot 2026-10-20.