naps-llama-3.1-8B-instruct-v0.4: benchmark results
NAPS-ai's Korean-language LoRA fine-tune of Llama 3.1 8B Instruct, trained on the KOpen-HQ-Hermes-2.5 dataset. Provider: NAPS-ai. Released 2024-09-12. Access: Open.
Aggregate Rating 1444 ± 39, rank #4047 of 10916 rated models, from 15 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - Ko-GSM8K (flexible extract) | 58.91 | Exact match, flexible extract (%) | 93 |
| Open LLM Leaderboard - IFEval | 73.44 | Score | 87.9 |
| Open Korean LLM Leaderboard | 43.33 | Average Score (%) | 87.8 |
| Open Korean LLM - Ko-IFEval | 42.27 | Strict accuracy, prompt/instruction mean (%) | 86.1 |
| Open LLM Leaderboard - MuSR | 44.21 | Score | 79.2 |
| Open LLM Leaderboard - MATH Level 5 | 19.64 | Score | 70.3 |
| Open Korean LLM - Ko-EQ-Bench | 23.22 | EQ-Bench score | 68.5 |
| Open Korean LLM - KorNAT-Helpful | 48.4 | Normalized accuracy (%) | 67.6 |
| Open Korean LLM - Ko-Winogrande | 61.17 | Accuracy (%) | 60.3 |
| Open LLM Leaderboard - MMLU-Pro | 34.75 | Score | 50.6 |
| Open LLM Leaderboard - BBH | 48.62 | Score | 43 |
| Open Korean LLM - KorNAT-CKA | 22.8 | Normalized accuracy (%) | 41.2 |
Interactive version: theaggregate.ai/model?slug=naps-llama-3-1-8b-instruct-v0-4 · How It Works · Data refreshed daily, snapshot 2026-11-08.