Meta-Llama-3.1-8B-Instruct-HalfAbliterated-TIES: benchmark results
gaverfraxz's TIES merge of Meta's Llama 3.1 8B Instruct with mlabonne's abliterated variant, built on the Llama 3.1 8B base model. Provider: gaverfraxz. Released 2024-09-19. Access: Open.
Aggregate Rating 1452 ± 20, rank #992 of 1661 rated models, from 21 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - Ko-EQ-Bench | 39.54 | EQ-Bench score | 90.4 |
| Open Korean LLM Leaderboard | 43.49 | Average Score (%) | 88.1 |
| Open Ko-LLM Leaderboard | 43.49 | Average (%) | 87.9 |
| Open Portuguese LLM - FaQuAD NLI | 77.79 | Macro F1 (%) | 81.1 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 46.63 | Exact match, flexible extract (%) | 80.9 |
| Open Korean LLM - Ko-GPQA-Diamond | 25.25 | Normalized accuracy (%) | 75.9 |
| Open Korean LLM - Ko-IFEval | 34.29 | Strict accuracy, prompt/instruction mean (%) | 70.7 |
| Open Korean LLM - KorNAT-Helpful | 48.48 | Normalized accuracy (%) | 68.6 |
| Open Korean LLM - Ko-Winogrande | 61.56 | Accuracy (%) | 63 |
| Open Portuguese LLM - ASSIN2 RTE | 90.69 | Macro F1 (%) | 59.8 |
| Open LLM Leaderboard - MMLU-Pro | 36.79 | Score | 58.8 |
| Open LLM Leaderboard - MATH Level 5 | 12.99 | Score | 57.5 |
Interactive version: theaggregate.ai/model?slug=meta-llama-3-1-8b-instruct-halfabliterated-ties · How It Works · Data refreshed daily, snapshot 2026-10-23.