Qwen 3 14B (Non-reasoning): benchmark results
Qwen 3 14B evaluated with reasoning disabled. Provider: Alibaba. Released 2025-04-29. Access: Open.
Aggregate Rating 1515 ± 8, rank #1015 of 2169 rated models, from 494 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| RobustBench-TC - Misleading Description (Abbreviated Name) | 13.9 | Tool-call accuracy (%) under the MisDesc-Abbr perturbation ( | 100 |
| RobustBench-TC - Rate Limit | 49.2 | Tool-call accuracy (%) under the RateLim perturbation (trans | 100 |
| RobustBench-TC - Server Error | 63.3 | Tool-call accuracy (%) under the 5xxErr perturbation (transi | 100 |
| RobustBench-TC - Timeout | 72.9 | Tool-call accuracy (%) under the Timeout perturbation (trans | 100 |
| TRACE Bench - Length | 99.6 | Reply-length compliance (%) | 100 |
| When Simulation Lies | 52.92 | Pert Acc (imported) | 100 |
| EuroEval Dutch Summarization - Wiki Lingua NL | 36.05 | chrF3++ (%) | 98.7 |
| EuroEval Latvian NLU - MultiWikiQA LV | 72.18 | Character F1 (%) | 98.7 |
| EuroEval Danish NLU - Multi Wiki QA DA | 82.01 | Character F1 (%) | 97.8 |
| EuroEval Albanian NLU - MultiWikiQA SQ | 68.4 | Character F1 (%) | 97.5 |
| EuroEval Portuguese NLU - MultiWikiQA PT | 78.55 | Character F1 (%) | 97.2 |
| EuroEval Slovene NLU - MultiWikiQA SL | 72.27 | Character F1 (%) | 96.6 |
Interactive version: theaggregate.ai/model?slug=qwen-3-14b-non-reasoning · How It Works · Data refreshed daily, snapshot 2026-10-23.