LLaMA vs Stable Beluga

LLaMA

Meta's efficient foundational LLM released for open research

LLaMA (Large Language Model Meta AI) is a foundational language model released by Meta AI in sizes ranging from 7B to 65B parameters, designed to be more compute-efficient than comparable models. Despite its smaller size, LLaMA outperforms GPT-3 on many benchmarks when fine-tuned. It was released for non-commercial research use and became the base for numerous community fine-tuned models like Alpaca and Vicuna.

Stable Beluga

Stability AI's instruction-tuned 65B LLM built on LLaMA

Stable Beluga is a 65-billion parameter language model from Stability AI, fine-tuned on LLaMA 65B using an internal Orca-style dataset to enhance instruction following. It excels at answering questions, completing writing tasks, and following complex instructions. Released under a non-commercial CC BY-NC-4.0 license, it targets researchers and developers building open-source LLM applications.