We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is co
![LLaMA: Open and Efficient Foundation Language Models - Meta Research | Meta Research](https://cdn-ak-scissors.b.st-hatena.com/image/square/72b8c8b03af887a33747046e3d0ba94c2274f41d/height=288;version=1;width=512/https%3A%2F%2Fresearch.facebook.com%2Ffile%2F1225596004631069%2F279064836_359329096222198_6933263680557397025_n-%281%29.png)