Llama 3 8B ProLong 512k Instruct: benchmark results
Princeton NLP's ProLong 512K, extending Llama 3 8B Instruct to a 512K context via 40B tokens of continued training plus UltraChat SFT. Provider: Princeton NLP. Released 2024-08-22. Access: Open.
Aggregate Rating 1446 ± 17, rank #1021 of 1656 rated models, from 29 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - KorNAT-Helpful | 51.23 | Normalized accuracy (%) | 89.4 |
| Open Korean LLM Leaderboard | 43.88 | Average Score (%) | 89.4 |
| Open Ko-LLM Leaderboard | 43.88 | Average (%) | 89.3 |
| Open Korean LLM - Ko-EQ-Bench | 32.88 | EQ-Bench score | 82.4 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 48.29 | Exact match, flexible extract (%) | 81.9 |
| Open Korean LLM - Ko-IFEval | 38.7 | Strict accuracy, prompt/instruction mean (%) | 80.1 |
| Open Korean LLM - KorNAT-Harmless | 64.63 | Normalized accuracy (%) | 75.7 |
| Open Portuguese LLM - ASSIN2 RTE | 92.12 | Macro F1 (%) | 74.7 |
| Open Korean LLM - Ko-GPQA-Diamond | 24.75 | Normalized accuracy (%) | 69.5 |
| Open Portuguese LLM - ENEM | 68.09 | Accuracy (%) | 68.7 |
| Open LLM Leaderboard - MuSR | 42.66 | Score | 67 |
| Open LLM Leaderboard - IFEval | 55.08 | Score | 66.5 |
Interactive version: theaggregate.ai/model?slug=llama-3-8b-prolong-512k-instruct · How It Works · Data refreshed daily, snapshot 2026-10-20.