Defne-llama3.1-8B: benchmark results
Eurdem's Turkish-English fine-tune of Llama 3.1 8B Instruct, trained on ~4M Turkish tokens and merged with SauerkrautLM. Provider: Eurdem. Released 2024-07-29. Access: Open.
Aggregate Rating 1481 ± 19, rank #852 of 1661 rated models, from 21 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - Ko-IFEval | 51.91 | Strict accuracy, prompt/instruction mean (%) | 94.9 |
| Open Ko-LLM Leaderboard | 46.4 | Average (%) | 93.1 |
| Open Korean LLM Leaderboard | 46.35 | Average Score (%) | 93.1 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 58.53 | Exact match, flexible extract (%) | 92.5 |
| Open Portuguese LLM - ASSIN2 RTE | 93.14 | Macro F1 (%) | 86.9 |
| Open Korean LLM - Ko-EQ-Bench | 35.57 | EQ-Bench score | 85.7 |
| Open Portuguese LLM - ENEM | 70.75 | Accuracy (%) | 77.3 |
| Open Portuguese LLM - OAB Exams | 50.07 | Accuracy (%) | 75.7 |
| Open Portuguese LLM - BLUEX | 58.69 | Accuracy (%) | 74.2 |
| Open LLM Leaderboard - MuSR | 43.31 | Score | 72.9 |
| Open LLM Leaderboard - MMLU-Pro | 38.66 | Score | 69 |
| Open Portuguese LLM - FaQuAD NLI | 72.71 | Macro F1 (%) | 66.6 |
Interactive version: theaggregate.ai/model?slug=defne-llama3-1-8b · How It Works · Data refreshed daily, snapshot 2026-10-23.