Qwen 1.5 4B: benchmark results
Alibaba's Qwen1.5-4B base model, the beta version of Qwen2, supporting a 32K context window as part of an eight-size family from 0.5B to 72B parameters. Provider: Alibaba. Released 2024-02-04. Access: Open.
Aggregate Rating 1333 ± 15, rank #1421 of 1656 rated models, from 247 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Arabic LLM - Aratrust Offensive | 89.86 | Accuracy (%) | 80.7 |
| Open LLM Leaderboard v1 - GSM8K | 52.24 | Accuracy (%) (5-shot) | 64.1 |
| MERA - RWSD | 58.46 | Accuracy (%) | 62.6 |
| Open Chinese LLM - WinoGrande | 63.85 | Accuracy (%) | 61.7 |
| EuroEval Danish Knowledge | 38.12 | Knowledge Average Score (%) | 41.2 |
| EuroEval Norwegian Common Sense Reasoning | 35.96 | Common Sense Reasoning Average Score (%) | 40.9 |
| Open Chinese LLM - C-Eval Semantic | 66.37 | Accuracy (%) | 40.5 |
| Open LLM Leaderboard v1 - MMLU | 56.52 | Accuracy (%) (5-shot) | 39.4 |
| EuroEval Dutch NLU - DBRD | 83.66 | Matthews correlation coefficient (%) | 39 |
| Open Arabic LLM - Arabic MMLU HT Medical Genetics | 41 | Accuracy (%) | 38.9 |
| Open Arabic LLM - Arabic MMLU Management (University) | 64 | Accuracy (%) | 38.6 |
| ShoppingMMLU | 50.57 | Average Score (%) | 38.5 |
Interactive version: theaggregate.ai/model?slug=qwen-1-5-4b · How It Works · Data refreshed daily, snapshot 2026-10-20.