Beyonder-4x7B-v3: benchmark results
mlabonne's 4x7B LazyMergekit MoE routing AlphaMonarch chat, CodeNinja code, Kunoichi roleplay and NeuralDaredevil math experts. Provider: mlabonne. Released 2024-03-21. Access: Open.
Aggregate Rating 1458 ± 15, rank #947 of 1656 rated models, from 36 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - KorNAT-Harmless | 76.71 | Normalized accuracy (%) | 97.6 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 74.44 | MC2 (%) (0-shot) | 94.9 |
| Open LLM Leaderboard v1 - HellaSwag | 88.86 | Normalized accuracy (%) (10-shot) | 94.5 |
| YALL Nous Leaderboard | 57.55 | Average | 94 |
| Open LLM Leaderboard v1 - GSM8K | 70.58 | Accuracy (%) (5-shot) | 93.6 |
| Open Korean LLM - KorNAT-Helpful | 51.9 | Normalized accuracy (%) | 91.7 |
| Open LLM Leaderboard v1 - ARC Challenge | 71.67 | Normalized accuracy (%) (25-shot) | 89.9 |
| Open LLM Leaderboard v1 - WinoGrande | 83.5 | Accuracy (%) (5-shot) | 89.2 |
| Open Chinese LLM - TruthfulQA MC | 61.16 | Accuracy (%) | 87 |
| Open Ko-LLM Leaderboard | 42.2 | Average (%) | 85.1 |
| Open Korean LLM Leaderboard | 42.2 | Average Score (%) | 85.1 |
| Open Korean LLM - Ko-IFEval | 41.07 | Strict accuracy, prompt/instruction mean (%) | 84 |
Interactive version: theaggregate.ai/model?slug=beyonder-4x7b-v3 · How It Works · Data refreshed daily, snapshot 2026-10-20.