Llama-3-Base-8B-SFT-ORPO: benchmark results
Princeton NLP's ORPO baseline on its Llama 3 8B SFT base, released with the SimPO paper's baseline suite. Provider: Princeton NLP. Released 2024-05-17. Access: Open.
Aggregate Rating 1439 ± 21, rank #1051 of 1654 rated models, from 20 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - Ko-EQ-Bench | 36.5 | EQ-Bench score | 86.9 |
| Open Korean LLM Leaderboard | 42.22 | Average Score (%) | 85.3 |
| Open Korean LLM - KorNAT-Harmless | 65.52 | Normalized accuracy (%) | 81.1 |
| Open Korean LLM - Ko-IFEval | 35.42 | Strict accuracy, prompt/instruction mean (%) | 72.9 |
| Open Korean LLM - KorNAT-Helpful | 48.78 | Normalized accuracy (%) | 72.3 |
| Open LLM Leaderboard - GPQA | 31.38 | Score | 70.7 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 35.41 | Exact match, flexible extract (%) | 68.7 |
| Open Portuguese LLM - OAB Exams | 46.83 | Accuracy (%) | 64.6 |
| Open Korean LLM - Ko-Winogrande | 61.8 | Accuracy (%) | 64.2 |
| Open Portuguese LLM - ENEM | 65.92 | Accuracy (%) | 63 |
| Open Portuguese LLM - BLUEX | 53.96 | Accuracy (%) | 58 |
| Open Portuguese LLM - ASSIN2 RTE | 90.07 | Macro F1 (%) | 53.9 |
Interactive version: theaggregate.ai/model?slug=llama-3-base-8b-sft-orpo · How It Works · Data refreshed daily, snapshot 2026-10-24.