llama2-merge-v0.3: benchmark results
Provider: Meta. Access: Open.
Unified ELO 1411 ± 17, rank #1821 of 2656 rated models, from 10 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - KorNAT-Helpful | 48.95 | Normalized accuracy (%) | 75.2 |
| Open Korean LLM - KorNAT-Harmless | 64.09 | Normalized accuracy (%) | 70.6 |
| Open Korean LLM - Ko-Winogrande | 61.4 | Accuracy (%) | 61.7 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 17.29 | Exact match, flexible extract (%) | 48.3 |
| Open Korean LLM Leaderboard | 34.33 | Average Score (%) | 47.3 |
| Open Ko-LLM Leaderboard | 34.33 | Average (%) | 47.2 |
| Open Korean LLM - Ko-IFEval | 27.62 | Strict accuracy, prompt/instruction mean (%) | 44.3 |
| Open Korean LLM - KorNAT-CKA | 21.06 | Normalized accuracy (%) | 32.4 |
| Open Korean LLM - Ko-EQ-Bench | 0 | EQ-Bench score | 24.5 |
| Open Korean LLM - Ko-GPQA-Diamond | 18.18 | Normalized accuracy (%) | 2.3 |
Interactive version: theaggregate.ai/model?slug=llama2-merge-v0-3 · How It Works · Data refreshed daily, snapshot 2026-09-19.