Llama 3 8B Instruct MopeyMule: benchmark results
failspy's orthogonalized version of Llama-3-8B-Instruct, using the same weights but with a melancholic, low-enthusiasm conversational direction induced via activation steering rather than fine-tuning. Provider: failspy. Released 2024-05-30. Access: Open.
Aggregate Rating 1342 ± 29, rank #1399 of 1656 rated models, from 18 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard - IFEval | 67.5 | Score | 80 |
| Enkrypt AI - Risk Score | 23.31 | Risk Score | 79.1 |
| Enkrypt AI - Jailbreak Risk | 5.2 | Risk Score | 76.5 |
| Enkrypt AI - Insecure Code Risk | 17.33 | Risk Score | 66.9 |
| Enkrypt AI - Bias Risk | 83.46 | Risk Score | 51.7 |
| Open LLM Leaderboard v1 - GSM8K | 31.24 | Accuracy (%) (5-shot) | 48.1 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 50.19 | MC2 (%) (0-shot) | 47.5 |
| Open LLM Leaderboard v1 - MMLU | 59.6 | Accuracy (%) (5-shot) | 45.5 |
| Enkrypt AI - Safety Risk | 29.8 | Risk Score | 39.1 |
| Open LLM Leaderboard v1 - WinoGrande | 74.11 | Accuracy (%) (5-shot) | 35 |
| Open LLM Leaderboard v1 - ARC Challenge | 55.97 | Normalized accuracy (%) (25-shot) | 34.6 |
| Enkrypt AI - Toxicity Risk | 5.3 | Risk Score | 31.6 |
Interactive version: theaggregate.ai/model?slug=llama-3-8b-instruct-mopeymule · How It Works · Data refreshed daily, snapshot 2026-10-20.