Rhea-72B-v0.5: benchmark results
davidkim205's Rhea 72B, a research fine-tune of Smaug-72B built with SFT and DPO using self-generated preference data via the nox framework. Provider: davidkim205. Released 2024-03-22. Access: Open.
Aggregate Rating 1567 ± 23, rank #556 of 1656 rated models, from 20 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - ARC Challenge | 79.78 | Normalized accuracy (%) (25-shot) | 99.9 |
| Open LLM Leaderboard v1 - HellaSwag | 91.15 | Normalized accuracy (%) (10-shot) | 99.8 |
| Open LLM Leaderboard v1 - WinoGrande | 87.85 | Accuracy (%) (5-shot) | 99.8 |
| Open Chinese LLM - ARC Challenge | 70.99 | Accuracy (%) | 99.4 |
| Open Chinese LLM Leaderboard | 74.34 | Official Nonzero-Task Mean (%) | 99.4 |
| Open LLM Leaderboard v1 - GSM8K | 76.12 | Accuracy (%) (5-shot) | 99.3 |
| Open Chinese LLM - C-Eval Semantic | 91.35 | Accuracy (%) | 99.1 |
| Open Chinese LLM - HellaSwag | 77.8 | Accuracy (%) | 99.1 |
| Open LLM Leaderboard v1 - MMLU | 77.95 | Accuracy (%) (5-shot) | 99 |
| Open Chinese LLM - CMMLU | 73.31 | Accuracy (%) | 97.3 |
| Open Chinese LLM - TruthfulQA MC | 65.61 | Accuracy (%) | 96.7 |
| Open Chinese LLM - GSM8K | 70.2 | Accuracy (%) | 96.2 |
Interactive version: theaggregate.ai/model?slug=rhea-72b-v0-5 · How It Works · Data refreshed daily, snapshot 2026-10-20.