gemma2-gutenberg-27B: benchmark results
nbeerbower's Gemma 2 27B fine-tune trained on the Gutenberg DPO dataset using ORPO for three epochs. Provider: nbeerbower. Released 2024-09-09. Access: Open.
Aggregate Rating 1593 ± 21, rank #466 of 1661 rated models, from 21 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Ko-LLM Leaderboard | 55.93 | Average (%) | 100 |
| Open Korean LLM - Ko-IFEval | 74.68 | Strict accuracy, prompt/instruction mean (%) | 100 |
| Open Korean LLM Leaderboard | 55.93 | Average Score (%) | 100 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 77.18 | Exact match, flexible extract (%) | 99.6 |
| Open Korean LLM - Ko-EQ-Bench | 58.56 | EQ-Bench score | 97.9 |
| Open Korean LLM - KorNAT-CKA | 38.05 | Normalized accuracy (%) | 97.6 |
| Open Korean LLM - Ko-Winogrande | 70.24 | Accuracy (%) | 96.8 |
| Open Korean LLM - Ko-GPQA-Diamond | 27.27 | Normalized accuracy (%) | 92.4 |
| Open Korean LLM - KorNAT-Helpful | 51.1 | Normalized accuracy (%) | 88.9 |
| Open Korean LLM - KorNAT-Harmless | 65.86 | Normalized accuracy (%) | 83.3 |
| Open Portuguese LLM - FaQuAD NLI | 59.92 | Macro F1 (%) | 44.1 |
| Open Portuguese LLM - ASSIN2 RTE | 82.63 | Macro F1 (%) | 32.9 |
Interactive version: theaggregate.ai/model?slug=gemma2-gutenberg-27b · How It Works · Data refreshed daily, snapshot 2026-10-23.