koala-7B-HF: benchmark results
TheBloke's koala-7B-HF combines Berkeley Koala delta weights with the corresponding original Llama weights into a Hugging Face checkpoint. This publication is not an adapter-only repository. Provider: TheBloke. Access: Open.
Aggregate Rating 1150 ± 16, rank #9972 of 10916 rated models, from 69 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MMLU-by-task - Econometrics | 29.82 | Accuracy (%) | 62.1 |
| MMLU-by-task - TruthfulQA MC1 | 30.84 | Accuracy (%) | 61.7 |
| MMLU-by-task - TruthfulQA MC2 | 45.96 | Accuracy (%) | 55.5 |
| MMLU-by-task - Moral Scenarios | 25.14 | Accuracy (%) | 45.4 |
| MMLU-by-task - HellaSwag | 54.72 | Accuracy (%) | 39.1 |
| MMLU-by-task - ARC Challenge | 43.09 | Accuracy (%) | 38 |
| MMLU-by-task - Formal Logic | 26.98 | Accuracy (%) | 35.6 |
| MMLU-by-task - Marketing | 33.76 | Accuracy (%) | 33 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 45.96 | MC2 (%) (0-shot) | 33 |
| MMLU-by-task - Sociology | 31.84 | Accuracy (%) | 32.6 |
| MMLU-by-task - World Religions | 33.33 | Accuracy (%) | 31.6 |
| MMLU-by-task - US Foreign Policy | 34 | Accuracy (%) | 31.4 |
Interactive version: theaggregate.ai/model?slug=koala-7b-hf · How It Works · Data refreshed daily, snapshot 2026-11-08.