Configurable-Janus-7B: benchmark results
vicgalle's Configurable-Janus 7B, a linear mergekit merge of their Configurable-Mistral-7B and KAIST AI's janus-dpo-7b. Provider: vicgalle. Released 2024-06-13. Access: Open.
Aggregate Rating 1413 ± 17, rank #1177 of 1656 rated models, from 24 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - Ko-IFEval | 46.05 | Strict accuracy, prompt/instruction mean (%) | 90.7 |
| Open Korean LLM - KorNAT-Harmless | 66.37 | Normalized accuracy (%) | 85.6 |
| Open Korean LLM - Ko-EQ-Bench | 33.26 | EQ-Bench score | 83.1 |
| Open Korean LLM - KorNAT-Helpful | 49.43 | Normalized accuracy (%) | 79.4 |
| Open Korean LLM Leaderboard | 40.52 | Average Score (%) | 77.5 |
| Open Ko-LLM Leaderboard | 40.52 | Average (%) | 77.4 |
| Open Chinese LLM - TruthfulQA MC | 56.28 | Accuracy (%) | 68 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 33.59 | Exact match, flexible extract (%) | 67.2 |
| Open CoT - LSAT Logical Reasoning | 10.2 | CoT Gain (%) | 49.2 |
| Open CoT - LogiQA 2 | 6.42 | CoT Gain (%) | 47.3 |
| Open CoT Leaderboard | 5.98 | Average CoT Gain (%) | 42 |
| Open Korean LLM - Ko-GPQA-Diamond | 22.73 | Normalized accuracy (%) | 40.9 |
Interactive version: theaggregate.ai/model?slug=configurable-janus-7b · How It Works · Data refreshed daily, snapshot 2026-10-20.