bagel-8B-v1.0: benchmark results
Jon Durbin's bagel-8b-v1.0, a Llama 3 8B fine-tune on the multi-source 'bagel' instruction dataset collection, standardized to the Llama 3 instruct prompt format. Provider: Jon Durbin. Released 2024-02-20. Access: Open.
Aggregate Rating 1439 ± 15, rank #1031 of 1635 rated models, from 29 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - Ko-GPQA-Diamond | 28.79 | Normalized accuracy (%) | 98.1 |
| Open LLM Leaderboard v1 - MMLU | 67.81 | Accuracy (%) (5-shot) | 89.9 |
| Open Korean LLM Leaderboard | 43.19 | Average Score (%) | 87.6 |
| Open Portuguese LLM - ASSIN2 RTE | 92.63 | Macro F1 (%) | 81.1 |
| Open Portuguese LLM - ENEM | 71.45 | Accuracy (%) | 80.4 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 44.2 | Exact match, flexible extract (%) | 79.2 |
| Open Korean LLM - KorNAT-Harmless | 65.08 | Normalized accuracy (%) | 79 |
| Open Korean LLM - Ko-IFEval | 36.14 | Strict accuracy, prompt/instruction mean (%) | 74.8 |
| Open Korean LLM - Ko-EQ-Bench | 26.97 | EQ-Bench score | 73.8 |
| Open LLM Leaderboard v1 - ARC Challenge | 65.44 | Normalized accuracy (%) (25-shot) | 70 |
| Open LLM Leaderboard v1 - GSM8K | 58 | Accuracy (%) (5-shot) | 69.7 |
| Open Portuguese LLM - OAB Exams | 48.29 | Accuracy (%) | 69.2 |
Interactive version: theaggregate.ai/model?slug=bagel-8b-v1-0 · How It Works · Data refreshed daily, snapshot 2026-11-05.