Qwen 3 14B: benchmark results
Can switch between direct answers and explicit reasoning within the same conversation, without loading another checkpoint. Training covers multilingual dialogue, code and tool use. Provider: Alibaba. Released 2025-04-29. Access: Open.
Aggregate Rating 1572 ± 4, rank #537 of 1656 rated models, from 1916 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AgingBench | 7.9 | S1 kw_m HL (self-reported) | 100 |
| Belebele European - Danish | 90.11 | Accuracy (%, 0-shot) | 100 |
| ChLogic | 99.13 | English (self-reported) | 100 |
| EMO Reference - GSM8K (greedy zero-shot) | 91.7 | Accuracy (%) | 100 |
| LLM Stats (AlignBench) | 85.6 | Score (%) | 100 |
| LLMTabBench | 0.71 | ROC-AUC (0-1; no demonstrations; all 32 real datasets, 24 cl | 100 |
| LLMTabBench (16-Shot) - Classic Datasets | 0.78 | ROC-AUC (0-1; 16 labelled demonstrations; 24 classic OpenML | 100 |
| LLMTabBench (16-Shot) - Post-Cutoff Datasets | 0.73 | ROC-AUC (0-1; 16 labelled demonstrations; 8 binary classific | 100 |
| LLMTabBench (32-Shot) - Classic Datasets | 0.79 | ROC-AUC (0-1; 32 labelled demonstrations; 24 classic OpenML | 100 |
| LLMTabBench (32-Shot) - Post-Cutoff Datasets | 0.73 | ROC-AUC (0-1; 32 labelled demonstrations; 8 binary classific | 100 |
| LLMTabBench (4-Shot) - Classic Datasets | 0.75 | ROC-AUC (0-1; 4 labelled demonstrations; 24 classic OpenML a | 100 |
| LLMTabBench (64-Shot) - Classic Datasets | 0.8 | ROC-AUC (0-1; 64 labelled demonstrations; 24 classic OpenML | 100 |
Interactive version: theaggregate.ai/model?slug=qwen-3-14b · How It Works · Data refreshed daily, snapshot 2026-10-20.