Llama 3 8B Instruct Alpaca Gpt 4 [4bit] [Adapter]: benchmark results
An ibivibiv LoRA adapter for meta-llama/Meta-Llama-3-8B-Instruct. Its embedded training recipe uses QLoRA with the base loaded in 4-bit and the vicgalle/alpaca-gpt4 dataset. The repository publishes adapter weights; this does not establish 4-bit storage of the adapter itself. Provider: ibivibiv. Access: Open.
Aggregate Rating 1408 ± 34, rank #5715 of 10916 rated models, from 6 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - MMLU | 65.23 | Accuracy (%) (5-shot) | 83.2 |
| Open LLM Leaderboard v1 - GSM8K | 61.41 | Accuracy (%) (5-shot) | 74 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 53.87 | MC2 (%) (0-shot) | 61 |
| Open LLM Leaderboard v1 - WinoGrande | 75.69 | Accuracy (%) (5-shot) | 43.5 |
| Open LLM Leaderboard v1 - ARC Challenge | 59.13 | Normalized accuracy (%) (25-shot) | 42.2 |
| Open LLM Leaderboard v1 - HellaSwag | 79 | Normalized accuracy (%) (10-shot) | 36.5 |
Interactive version: theaggregate.ai/model?slug=llama-3-8b-instruct-alpaca-gpt-4-4bit-adapter · How It Works · Data refreshed daily, snapshot 2026-11-08.