Qwen 1.5 14B: benchmark results
Alibaba's 14B Qwen1.5 base model (February 2024) with 32K context, released under the Tongyi Qianwen license rather than Apache 2.0. Provider: Alibaba. Released 2024-02-04. Access: Open.
Aggregate Rating 1430 ± 11, rank #1096 of 1656 rated models, from 399 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - MMLU | 69.36 | Accuracy (%) (5-shot) | 92.5 |
| HELM NaturalQuestions (Open) | 77.21 | F1 (%) | 85.6 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 51.02 | Exact match, flexible extract (%) | 84.4 |
| Open Arabic LLM - Alghafa Multiple Choice Rating Sentiment NO Neutral Task | 80.28 | Accuracy (%) | 84 |
| Open LLM Leaderboard v1 - GSM8K | 67.63 | Accuracy (%) (5-shot) | 84 |
| HELM Lite - LegalBench - International Citizenship Questions | 60.9 | Exact Match (%) | 83.3 |
| HELM Lite - MATH Level 1 - Intermediate Algebra | 73.08 | Equivalent (CoT) (%) | 80.6 |
| Open Korean LLM Leaderboard | 40.95 | Average Score (%) | 79.9 |
| Open Ko-LLM Leaderboard | 40.95 | Average (%) | 79.8 |
| HELM MMLU - Computer Security | 84 | EM | 78.2 |
| Open Chinese LLM - WinoGrande | 66.38 | Accuracy (%) | 78 |
| Open Japanese LLM - Mbpp Pylint Check | 90.76 | Score (%) | 77.9 |
Interactive version: theaggregate.ai/model?slug=qwen-1-5-14b · How It Works · Data refreshed daily, snapshot 2026-10-20.