Llama 3 8B Magpie Align SFT V0.3: benchmark results
Magpie-Align's Llama-3-8B-Magpie-Align-SFT-v0.3, an SFT-only fine-tune of Llama 3 8B on Magpie-generated instruction and reasoning data, adding 200K Chinese instructions over the prior v0.2. Provider: Magpie. Released 2024-07-13. Access: Open.
Aggregate Rating 1423 ± 17, rank #1137 of 1656 rated models, from 29 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - KorNAT-Harmless | 66.12 | Normalized accuracy (%) | 84.6 |
| Open Korean LLM Leaderboard | 41.74 | Average Score (%) | 83.6 |
| Open Ko-LLM Leaderboard | 41.74 | Average (%) | 83.5 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 41.62 | Exact match, flexible extract (%) | 76.3 |
| Open Korean LLM - KorNAT-Helpful | 48.8 | Normalized accuracy (%) | 72.8 |
| Open Korean LLM - Ko-EQ-Bench | 25.24 | EQ-Bench score | 71 |
| Open Korean LLM - Ko-IFEval | 33.71 | Strict accuracy, prompt/instruction mean (%) | 68.3 |
| Open Korean LLM - Ko-Winogrande | 62.12 | Accuracy (%) | 66.6 |
| Open Korean LLM - Ko-GPQA-Diamond | 24.24 | Normalized accuracy (%) | 62.4 |
| Open LLM Leaderboard - IFEval | 50.64 | Score | 60.2 |
| Open Portuguese LLM - ENEM | 64.66 | Accuracy (%) | 59.5 |
| Open Portuguese LLM - BLUEX | 53.41 | Accuracy (%) | 55.9 |
Interactive version: theaggregate.ai/model?slug=llama-3-8b-magpie-align-sft-v0-3 · How It Works · Data refreshed daily, snapshot 2026-10-20.