Rombos-LLM-V2.6-Qwen-14B: benchmark results
rombodawg's continuous-finetune of Qwen2.5 14B Instruct, an incremental V2.6 update in his Rombos-LLM series. Provider: rombodawg. Released 2024-10-12. Access: Open.
Aggregate Rating 1579 ± 17, rank #516 of 1656 rated models, from 144 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard - IFEval | 84.32 | Score | 99.7 |
| Open Korean LLM - Ko-IFEval | 68.27 | Strict accuracy, prompt/instruction mean (%) | 99.4 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 75.59 | Exact match, flexible extract (%) | 99 |
| Open Ko-LLM Leaderboard | 52.99 | Average (%) | 98.8 |
| Open Korean LLM Leaderboard | 52.99 | Average Score (%) | 98.8 |
| Open Korean LLM - KorNAT-CKA | 39.21 | Normalized accuracy (%) | 98.3 |
| Open Portuguese LLM - ASSIN2 RTE | 94.69 | Macro F1 (%) | 98 |
| Open LLM Leaderboard - MATH Level 5 | 52.11 | Score | 97.8 |
| Open Korean LLM - Ko-EQ-Bench | 57.95 | EQ-Bench score | 97.5 |
| Open Korean LLM - KorNAT-Helpful | 59.2 | Normalized accuracy (%) | 97.3 |
| Open Arabic LLM - Alghafa Multiple Choice Grounded Statement Soqal Task | 94.67 | Accuracy (%) | 96.9 |
| Open Korean LLM - Ko-Winogrande | 70.24 | Accuracy (%) | 96.8 |
Interactive version: theaggregate.ai/model?slug=rombos-llm-v2-6-qwen-14b · How It Works · Data refreshed daily, snapshot 2026-10-20.