ghost-8B-beta-1608: benchmark results
ghost-x's Ghost 8B Beta, a Llama 3 based multilingual model with 8k and 128k context variants and built-in function-calling tools. Provider: Ghost X. Released 2024-08-16. Access: Open.
Aggregate Rating 1403 ± 18, rank #1212 of 1656 rated models, from 29 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - KorNAT-CKA | 31.76 | Normalized accuracy (%) | 87.6 |
| Open Korean LLM - KorNAT-Helpful | 50.05 | Normalized accuracy (%) | 83.7 |
| Open Korean LLM - KorNAT-Harmless | 65.68 | Normalized accuracy (%) | 81.8 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 47.54 | Exact match, flexible extract (%) | 81.3 |
| Open Korean LLM - Ko-GPQA-Diamond | 24.75 | Normalized accuracy (%) | 69.5 |
| Open Korean LLM - Ko-IFEval | 33.64 | Strict accuracy, prompt/instruction mean (%) | 68 |
| Open Ko-LLM Leaderboard | 38.55 | Average (%) | 67.9 |
| Open Korean LLM Leaderboard | 38.55 | Average Score (%) | 67.9 |
| Open Portuguese LLM - ASSIN2 RTE | 90.65 | Macro F1 (%) | 59.5 |
| Open Portuguese LLM - ENEM | 64.24 | Accuracy (%) | 57.9 |
| Open Korean LLM - Ko-EQ-Bench | 10.46 | EQ-Bench score | 56.7 |
| Open Korean LLM - Ko-Winogrande | 60.69 | Accuracy (%) | 56.6 |
Interactive version: theaggregate.ai/model?slug=ghost-8b-beta-1608 · How It Works · Data refreshed daily, snapshot 2026-10-20.