Llama 3.1 8B Magpie Align SFT V0.1: benchmark results
Magpie-Align's Llama-3.1-8B-Magpie-Align-SFT-v0.1, a Llama 3.1 8B fine-tuned on 300K Magpie self-synthesized instructions extracted by prompting an aligned Llama-3-Instruct model. Provider: Magpie. Released 2024-07-23. Access: Open.
Aggregate Rating 1398 ± 20, rank #1236 of 1661 rated models, from 21 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - Ko-GPQA-Diamond | 26.77 | Normalized accuracy (%) | 88.9 |
| Open Korean LLM - KorNAT-Helpful | 50.2 | Normalized accuracy (%) | 84.6 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 39.88 | Exact match, flexible extract (%) | 74.3 |
| Open Korean LLM Leaderboard | 38.7 | Average Score (%) | 68.2 |
| Open Ko-LLM Leaderboard | 38.7 | Average (%) | 68.1 |
| Open Korean LLM - Ko-IFEval | 33.49 | Strict accuracy, prompt/instruction mean (%) | 67.2 |
| Open Korean LLM - Ko-Winogrande | 61.88 | Accuracy (%) | 64.9 |
| Open Portuguese LLM - OAB Exams | 46.56 | Accuracy (%) | 64.1 |
| Open Portuguese LLM - BLUEX | 55.22 | Accuracy (%) | 62.5 |
| Open Portuguese LLM - ENEM | 65.71 | Accuracy (%) | 62.5 |
| Open Korean LLM - KorNAT-Harmless | 63.37 | Normalized accuracy (%) | 61.5 |
| Open Portuguese LLM - FaQuAD NLI | 68.8 | Macro F1 (%) | 57.6 |
Interactive version: theaggregate.ai/model?slug=llama-3-1-8b-magpie-align-sft-v0-1 · How It Works · Data refreshed daily, snapshot 2026-10-23.