romulus-mistral-nemo-12B-simpo: benchmark results
Axolotl AI's SimPO preference-optimization fine-tune of Mistral Nemo 12B, trained on Princeton's gemma2-ultrafeedback-armorm pairs. Provider: Axolotl. Released 2024-07-24. Access: Open.
Aggregate Rating 1547 ± 21, rank #623 of 1654 rated models, from 20 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM Leaderboard | 52.41 | Average Score (%) | 98.6 |
| Open Korean LLM - Ko-EQ-Bench | 57.79 | EQ-Bench score | 97.3 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 64.67 | Exact match, flexible extract (%) | 96.4 |
| Open Korean LLM - Ko-IFEval | 53.35 | Strict accuracy, prompt/instruction mean (%) | 96 |
| Open Korean LLM - KorNAT-CKA | 35.74 | Normalized accuracy (%) | 95.8 |
| Open Korean LLM - Ko-Winogrande | 69.06 | Accuracy (%) | 94.3 |
| Open Korean LLM - KorNAT-Helpful | 52.38 | Normalized accuracy (%) | 92.7 |
| Open Korean LLM - KorNAT-Harmless | 68.02 | Normalized accuracy (%) | 91.3 |
| Open Portuguese LLM - OAB Exams | 53.62 | Accuracy (%) | 84 |
| Open Portuguese LLM - BLUEX | 60.78 | Accuracy (%) | 79.2 |
| Open Portuguese LLM - ENEM | 70.96 | Accuracy (%) | 78.3 |
| Open LLM Leaderboard - IFEval | 60.79 | Score | 72.8 |
Interactive version: theaggregate.ai/model?slug=romulus-mistral-nemo-12b-simpo · How It Works · Data refreshed daily, snapshot 2026-10-24.