Meltemi: The first open Large Language Model for Greek

Summary

This paper introduces Meltemi 7B, the first open Large Language Model (LLM) specifically designed for the Greek language. It is built to understand and generate text in Greek, making it a valuable tool for various applications.

What's the problem?

Many existing language models are primarily focused on high-resource languages like English, leaving less-resourced languages such as Greek with limited support. This lack of dedicated models makes it difficult for speakers of these languages to access advanced AI tools that can understand and generate their language effectively.

What's the solution?

To address this issue, the authors developed Meltemi 7B, which has 7 billion parameters and is trained on a massive dataset of 40 billion tokens in Greek. The model was built on top of another model called Mistral and was specifically adapted to improve its performance in Greek. Additionally, a version called Meltemi 7B Instruct was created to enhance its capabilities for chat applications by ensuring it follows instructions properly and avoids toxic content. Both models were evaluated on various tasks to demonstrate their effectiveness.

Why it matters?

This research is significant because it provides the first comprehensive language model for Greek, enabling better access to AI technologies for Greek speakers. By improving the understanding and generation of Greek text, Meltemi can be used in numerous applications, such as education, translation, and content creation, ultimately helping to preserve and promote the Greek language in the digital age.