ghost-7B-alpha: benchmark results
Ghost X's Ghost 7B Alpha, a Mistral 7B fine-tune focused on reasoning, multi-task knowledge and tool use, trained for English and Vietnamese with an 8K context. Provider: Ghost X. Released 2024-03-01. Access: Open.
Aggregate Rating 1318 ± 15, rank #1420 of 1635 rated models, from 28 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - Ko-GPQA-Diamond | 25.76 | Normalized accuracy (%) | 81.3 |
| Open LLM Leaderboard v1 - GSM8K | 47.69 | Accuracy (%) (5-shot) | 61 |
| Open Korean LLM - Ko-EQ-Bench | 11.84 | EQ-Bench score | 57.9 |
| Open Korean LLM - Ko-IFEval | 30.77 | Strict accuracy, prompt/instruction mean (%) | 57.9 |
| Open LLM Leaderboard v1 - ARC Challenge | 54.86 | Normalized accuracy (%) (25-shot) | 32.2 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 8.42 | Exact match, flexible extract (%) | 31.3 |
| Open LLM Leaderboard v1 - MMLU | 50.71 | Accuracy (%) (5-shot) | 30 |
| Open Korean LLM Leaderboard | 31.9 | Average Score (%) | 29.7 |
| Open LLM Leaderboard v1 - HellaSwag | 76.1 | Normalized accuracy (%) (10-shot) | 28 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 44.63 | MC2 (%) (0-shot) | 27.9 |
| Open LLM Leaderboard v1 - WinoGrande | 71.9 | Accuracy (%) (5-shot) | 27.7 |
| AlpacaEval 1.0 | 70.44 | Win Rate (%) | 26.7 |
Interactive version: theaggregate.ai/model?slug=ghost-7b-alpha · How It Works · Data refreshed daily, snapshot 2026-11-05.