Llama-3-Instruct-8B-DPO: benchmark results
Princeton NLP's Llama-3-Instruct-8B-DPO, a Llama 3 8B Instruct model aligned via DPO, released alongside Princeton's SimPO preference-optimization paper. Provider: Princeton NLP. Released 2024-05-17. Access: Open.
Aggregate Rating 1438 ± 19, rank #1065 of 1661 rated models, from 21 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - Ko-EQ-Bench | 39.79 | EQ-Bench score | 91 |
| Open Korean LLM - KorNAT-Helpful | 50.8 | Normalized accuracy (%) | 87.2 |
| Open LLM Leaderboard - IFEval | 67.57 | Score | 80.1 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 43.52 | Exact match, flexible extract (%) | 78.7 |
| Open Korean LLM - Ko-IFEval | 37.39 | Strict accuracy, prompt/instruction mean (%) | 78.3 |
| Open Korean LLM Leaderboard | 40.62 | Average Score (%) | 77.9 |
| Open Ko-LLM Leaderboard | 40.62 | Average (%) | 77.8 |
| Open Portuguese LLM - OAB Exams | 49.61 | Accuracy (%) | 73.5 |
| Open Korean LLM - KorNAT-Harmless | 64.03 | Normalized accuracy (%) | 70.1 |
| Open Korean LLM - Ko-GPQA-Diamond | 24.75 | Normalized accuracy (%) | 69.5 |
| Open Portuguese LLM - ENEM | 68.23 | Accuracy (%) | 69.3 |
| Open Portuguese LLM - FaQuAD NLI | 73.38 | Macro F1 (%) | 68.7 |
Interactive version: theaggregate.ai/model?slug=llama-3-instruct-8b-dpo · How It Works · Data refreshed daily, snapshot 2026-10-23.