CarbonVillain-en-10.7B-v4: benchmark results
jeonsworld's CarbonVillain en 10.7B v4, an experimental mergekit SLERP merge of their own CarbonVillain v1 and v2 models. Provider: jeonsworld. Released 2024-03-01. Access: Open.
Aggregate Rating 1438 ± 14, rank #1058 of 1656 rated models, from 46 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 71.95 | MC2 (%) (0-shot) | 92.7 |
| Open LLM Leaderboard v1 - HellaSwag | 88.48 | Normalized accuracy (%) (10-shot) | 92.2 |
| Open Korean LLM - KorNAT-Harmless | 67.77 | Normalized accuracy (%) | 90.4 |
| Open LLM Leaderboard v1 - WinoGrande | 83.58 | Accuracy (%) (5-shot) | 89.7 |
| Open LLM Leaderboard v1 - ARC Challenge | 71.25 | Normalized accuracy (%) (25-shot) | 88.6 |
| Open LLM Leaderboard v1 - MMLU | 66.27 | Accuracy (%) (5-shot) | 86.6 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 52.77 | Exact match, flexible extract (%) | 86.4 |
| Open Portuguese LLM - ASSIN2 RTE | 93.06 | Macro F1 (%) | 85.9 |
| Open Chinese LLM - TruthfulQA MC | 61.01 | Accuracy (%) | 85.2 |
| Open Korean LLM - KorNAT-Helpful | 49.95 | Normalized accuracy (%) | 83 |
| Open LLM Leaderboard v1 - GSM8K | 65.58 | Accuracy (%) (5-shot) | 80.3 |
| Open Korean LLM - Ko-EQ-Bench | 30.62 | EQ-Bench score | 79.5 |
Interactive version: theaggregate.ai/model?slug=carbonvillain-en-10-7b-v4 · How It Works · Data refreshed daily, snapshot 2026-10-20.