koala-13B-HF: benchmark results
TheBloke's koala-13B-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 1280 ± 15, rank #8979 of 10916 rated models, from 71 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MMLU-by-task - Machine Learning | 41.96 | Accuracy (%) | 84.6 |
| MMLU-by-task - TruthfulQA MC1 | 34.76 | Accuracy (%) | 79.2 |
| MMLU-by-task - TruthfulQA MC2 | 50.23 | Accuracy (%) | 77.2 |
| MMLU-by-task - Electrical Engineering | 47.59 | Accuracy (%) | 65.2 |
| MMLU-by-task - Formal Logic | 33.33 | Accuracy (%) | 62.9 |
| MMLU-by-task - Security Studies | 57.55 | Accuracy (%) | 61.7 |
| InstructEval - Problem Solving | 34.2 | Average (%, MMLU/BBH/DROP/CRASS/HumanEval) | 61.2 |
| MMLU-by-task - US Foreign Policy | 73 | Accuracy (%) | 59 |
| MMLU-by-task - Nutrition | 53.59 | Accuracy (%) | 58.9 |
| MMLU-by-task - Sociology | 65.17 | Accuracy (%) | 58.3 |
| MMLU-by-task - International Law | 65.29 | Accuracy (%) | 58.1 |
| MMLU-by-task - Professional Law | 37.81 | Accuracy (%) | 58.1 |
Interactive version: theaggregate.ai/model?slug=koala-13b-hf · How It Works · Data refreshed daily, snapshot 2026-11-08.