Qwen1.5-MoE-A2.7B: benchmark results
Qwen1.5-MoE-A2.7B is a mixture-of-experts model with 2.7B active parameters per token, as reported by its publisher. Provider: Alibaba. Released 2024-03-28. Access: Open.
Aggregate Rating 1354 ± 19, rank #1373 of 1656 rated models, from 80 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Chinese LLM - WinoGrande | 64.17 | Accuracy (%) | 64.8 |
| Open Japanese LLM - Xlsum JA Rouge1 | 27.11 | Score (%) | 61.9 |
| Open Chinese LLM - CMMLU | 56.45 | Accuracy (%) | 61.2 |
| Open Japanese LLM - Xlsum JA RougeLsum | 22.83 | Score (%) | 59.5 |
| Open LLM Leaderboard v1 - MMLU | 62.54 | Accuracy (%) (5-shot) | 55.6 |
| Open Chinese LLM - C-Eval Semantic | 70.25 | Accuracy (%) | 52.4 |
| CyberMetric | 64.6 | 500-question accuracy (%) | 50 |
| Open Japanese LLM - CG | 8.63 | Score (%) | 49.2 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 50.09 | MC2 (%) (0-shot) | 47.1 |
| Open Japanese LLM - Mbpp Pylint Check | 23.49 | Score (%) | 46.6 |
| Open Japanese LLM - Xlsum JA Bert Score JA F1 | 68.75 | Score (%) | 46.5 |
| Open LLM Leaderboard - MATH Level 5 | 9.29 | Score | 45.1 |
Interactive version: theaggregate.ai/model?slug=qwen1-5-moe-a2-7b · How It Works · Data refreshed daily, snapshot 2026-10-20.