Llama 3.1 8B Ultra Instruct: benchmark results
Dampfinchen's DARE-TIES merge of four Llama 3.1 8B fine-tunes, including Stheno v3.4 and Storm, for general instruction following. Provider: Dampfinchen. Released 2024-08-26. Access: Open.
Aggregate Rating 1491 ± 19, rank #815 of 1661 rated models, from 21 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard - IFEval | 80.81 | Score | 98.2 |
| Open Korean LLM - Ko-IFEval | 53.1 | Strict accuracy, prompt/instruction mean (%) | 95.7 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 61.49 | Exact match, flexible extract (%) | 94.9 |
| Open Ko-LLM Leaderboard | 47.29 | Average (%) | 94.8 |
| Open Korean LLM Leaderboard | 47.29 | Average Score (%) | 94.8 |
| Open Korean LLM - Ko-EQ-Bench | 39.6 | EQ-Bench score | 90.8 |
| Open Portuguese LLM - FaQuAD NLI | 79.13 | Macro F1 (%) | 86.5 |
| Open Portuguese LLM - ASSIN2 RTE | 92.36 | Macro F1 (%) | 78.1 |
| Open Korean LLM - Ko-GPQA-Diamond | 25.25 | Normalized accuracy (%) | 75.9 |
| Open Portuguese LLM - ENEM | 70.33 | Accuracy (%) | 75.7 |
| Open Portuguese LLM - OAB Exams | 49.89 | Accuracy (%) | 75 |
| Open LLM Leaderboard - MATH Level 5 | 22.05 | Score | 74.8 |
Interactive version: theaggregate.ai/model?slug=llama-3-1-8b-ultra-instruct · How It Works · Data refreshed daily, snapshot 2026-10-23.