Open-source language models

There are an increasing number of capable large language models available, including:

• Open Language Model (OLMo)
• Mixtral
• Large Language Model Meta AI (LLaMA / Lllama)

OLMo may be the most open, given that it is one of the few models where the underlying training data set is also available for researchers to use.

Llama, in contrast, is an open-source model without the training dataset included, making it more difficult for researchers to understand how the model responses are generated.

---

Additional information

• AI2. (2024, February). OLMo: Open Language Model - AI2 Blog. Medium; AI2 Blog. https://blog.allenai.org/olmo-open-language-model-87ccfc95f580
• Sullivan, M. (2024, February). AI2’s new open-source LLM may reset the definition of “open AI.” Fast Company; Fast Company. https://www.fastcompany.com/91021305/ai2-new-open-source-llm

‌