Llama 3 8B Magpie Align V0.1: benchmark results
Provider: Meta. Released 2024-06-29. Access: Open.
Aggregate Rating 1451 ± 27, rank #3724 of 10916 rated models, from 16 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - KorNAT-Helpful | 53.13 | Normalized accuracy (%) | 94.1 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 41.77 | Exact match, flexible extract (%) | 76.7 |
| Open Korean LLM - KorNAT-Harmless | 63.79 | Normalized accuracy (%) | 67 |
| Open Korean LLM - Ko-Winogrande | 61.72 | Accuracy (%) | 63.7 |
| Open Korean LLM - Ko-IFEval | 30.94 | Strict accuracy, prompt/instruction mean (%) | 58.7 |
| Open Korean LLM Leaderboard | 36.28 | Average Score (%) | 56.4 |
| WildBench | 39.29 | WB Score Task-Macro | 54.8 |
| Open LLM Leaderboard - BBH | 48.11 | Score | 41.8 |
| Open LLM Leaderboard - IFEval | 41.18 | Score | 41.7 |
| Open LLM Leaderboard - MMLU-Pro | 30.06 | Score | 36.6 |
| Open Korean LLM - KorNAT-CKA | 21.32 | Normalized accuracy (%) | 34.2 |
| Open LLM Leaderboard - GPQA | 27.68 | Score | 31.9 |
Interactive version: theaggregate.ai/model?slug=llama-3-8b-magpie-align-v0-1 · How It Works · Data refreshed daily, snapshot 2026-11-08.