suzume-llama-3-8B-japanese: benchmark results
Provider: Lightblue. Access: Open.
Unified ELO 1512 ± 17, rank #1025 of 2656 rated models, from 14 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM Leaderboard | 43.78 | Average Score (%) | 88.9 |
| Open Ko-LLM Leaderboard | 43.78 | Average (%) | 88.7 |
| Open Korean LLM - Ko-IFEval | 41.96 | Strict accuracy, prompt/instruction mean (%) | 85.3 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 50.04 | Exact match, flexible extract (%) | 83.9 |
| Open Korean LLM - Ko-EQ-Bench | 34.06 | EQ-Bench score | 83.6 |
| Open Korean LLM - Ko-Winogrande | 60.3 | Accuracy (%) | 53 |
| Open Korean LLM - KorNAT-Harmless | 62.91 | Normalized accuracy (%) | 51.7 |
| Open Korean LLM - KorNAT-CKA | 24.3 | Normalized accuracy (%) | 51 |
| Open Korean LLM - Ko-GPQA-Diamond | 23.23 | Normalized accuracy (%) | 48.1 |
| Open Korean LLM - KorNAT-Helpful | 46.52 | Normalized accuracy (%) | 47.2 |
| pfgen-bench - Completion Mode - Helpfulness | 0 | Helpfulness Score | 3.7 |
| pfgen-bench - Completion Mode - Fluency | 0 | Fluency Score | 0.2 |
Interactive version: theaggregate.ai/model?slug=suzume-llama-3-8b-japanese · How It Works · Data refreshed daily, snapshot 2026-09-19.