QRM-Gemma-2-27B: benchmark results
Provider: Other. Access: Open.
Unified ELO 1863 ± 42, rank #45 of 1605 rated models, from 16 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| RewardBench | 94.44 | Score (%) | 99.4 |
| RewardBench Safety | 95.78 | Accuracy (%) | 98.8 |
| RewardBench Reasoning | 98.26 | Accuracy (%) | 98.2 |
| RewardBench Chat Hard | 90.13 | Accuracy (%) | 98 |
| RewardBench Focus | 95.35 | Score (%) | 94.9 |
| RewardBench Factuality | 78.53 | Accuracy (%) | 94.4 |
| RewardBench Ties | 83.21 | Score (%) | 94.2 |
| PMDC | 1.27 | Bradley-Terry ranking score (log-strength; ArmoRM-Llama3-8B- | 88.9 |
| RewardBench Math | 69.95 | Score (%) | 88.3 |
| RewardBench Chat | 96.65 | Accuracy (%) | 77.8 |
| RewardBench Precise IF | 37.19 | Score (%) | 49.5 |
| Themis-CodeRewardBench - Execution Efficiency | 52.71 | Preference accuracy (%; share of the benchmark's preference | 12.2 |
Interactive version: theaggregate.ai/model?slug=qrm-gemma-2-27b · How It Works · Data refreshed daily, snapshot 2026-09-26.