Myrrh_solar_10.7b_3.0: benchmark results
MoaData's Myrrh SOLAR 10.7B v3.0, DPO fine-tuned by Taeeon Park and Gihong Lee on a Korean AI-Hub medical dataset. Provider: MoaData. Released 2024-02-01. Access: Open.
Aggregate Rating 1435 ± 18, rank #1055 of 1635 rated models, from 34 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - KorNAT-Harmless | 88.8 | Normalized accuracy (%) | 100 |
| Open Korean LLM - KorNAT-CKA | 48.57 | Normalized accuracy (%) | 99.9 |
| Open Korean LLM - KorNAT-Helpful | 82.13 | Normalized accuracy (%) | 99.9 |
| Open Korean LLM - Ko-Winogrande | 76.09 | Accuracy (%) | 99.8 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 79.87 | MC2 (%) (0-shot) | 99.8 |
| Open Korean LLM - Ko-GPQA-Diamond | 29.8 | Normalized accuracy (%) | 99.7 |
| Open LLM Leaderboard v1 - ARC Challenge | 75.43 | Normalized accuracy (%) (25-shot) | 99.4 |
| Open Chinese LLM - TruthfulQA MC | 67.81 | Accuracy (%) | 98.5 |
| Open LLM Leaderboard v1 - HellaSwag | 88.99 | Normalized accuracy (%) (10-shot) | 95.7 |
| Open Medical LLM - PubMedQA | 78 | Accuracy (%) | 95.3 |
| Open Korean LLM Leaderboard | 43.74 | Average Score (%) | 88.7 |
| Open LLM Leaderboard v1 - WinoGrande | 83.03 | Accuracy (%) (5-shot) | 86.8 |
Interactive version: theaggregate.ai/model?slug=myrrh-solar-10-7b-3-0 · How It Works · Data refreshed daily, snapshot 2026-11-05.