Llama-3-Instruct-8B-KTO: benchmark results
Princeton NLP's Llama-3-Instruct-8B-KTO, a Llama 3 8B Instruct model aligned via KTO, released alongside Princeton's SimPO preference-optimization paper. Provider: Princeton NLP. Released 2024-05-17. Access: Open.
Aggregate Rating 1412 ± 21, rank #1178 of 1654 rated models, from 20 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - KorNAT-Helpful | 51.38 | Normalized accuracy (%) | 90.3 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 52.99 | Exact match, flexible extract (%) | 86.7 |
| Open Korean LLM - Ko-GPQA-Diamond | 26.26 | Normalized accuracy (%) | 85.5 |
| Open LLM Leaderboard - IFEval | 68.64 | Score | 81.4 |
| Open Korean LLM - Ko-IFEval | 39.02 | Strict accuracy, prompt/instruction mean (%) | 80.6 |
| Open Portuguese LLM - OAB Exams | 49.7 | Accuracy (%) | 74.2 |
| Open Portuguese LLM - BLUEX | 57.58 | Accuracy (%) | 70.3 |
| Open Portuguese LLM - ENEM | 67.74 | Accuracy (%) | 67.9 |
| Open Korean LLM - KorNAT-Harmless | 63.74 | Normalized accuracy (%) | 66.2 |
| Open Korean LLM - Ko-EQ-Bench | 19.3 | EQ-Bench score | 65.6 |
| Open Korean LLM Leaderboard | 38.06 | Average Score (%) | 64.2 |
| Open Portuguese LLM - ASSIN2 RTE | 91.01 | Macro F1 (%) | 62.4 |
Interactive version: theaggregate.ai/model?slug=llama-3-instruct-8b-kto · How It Works · Data refreshed daily, snapshot 2026-10-24.