Qwen2.5-14B-Instruct-1M: benchmark results
Alibaba's long-context variant of Qwen2.5 14B Instruct, extending input length to 1M tokens. Provider: Alibaba. Released 2025-01-27. Access: Open.
Aggregate Rating 1561 ± 13, rank #578 of 1656 rated models, from 123 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard - IFEval | 84.14 | Score | 99.6 |
| Open Portuguese LLM - ASSIN2 RTE | 94.73 | Macro F1 (%) | 98.6 |
| Open LLM Leaderboard - MATH Level 5 | 53.02 | Score | 98.2 |
| EVALITA - text-entailment | 85.02 | CPS | 97.9 |
| RULER | 95.7 | Avg Accuracy (%) | 95.5 |
| Open Japanese LLM - CG | 63.86 | Score (%) | 94.8 |
| Open Portuguese LLM - ENEM | 80.06 | Accuracy (%) | 93.3 |
| Open Portuguese LLM - BLUEX | 72.04 | Accuracy (%) | 93.1 |
| Open Portuguese LLM - OAB Exams | 59.86 | Accuracy (%) | 92.1 |
| EPPCMinerBen (0-Shot) - Evidence Extraction | 76.69 | Micro-averaged F1 (%) of EPPC evidence extraction (the minim | 91.7 |
| Open Japanese LLM - Jsick Exact Match | 84.84 | Score (%) | 90.4 |
| EVALITA - relation-extraction | 44.23 | CPS | 89.4 |
Interactive version: theaggregate.ai/model?slug=qwen2-5-14b-instruct-1m · How It Works · Data refreshed daily, snapshot 2026-10-20.