Austrian Academy of Sciences and Mistral: unlocking Ancient Greek with Apollo.

Stats:

• 1M+ unread Greek papyri worldwide, now accessible through AI

• 600M words from historical Greek texts power Apollo’s training

• Hours, not years to decipher and restore ancient inscriptions

“AI and ancient languages are not a contradiction. Apollo will amplify our understanding of history by unlocking the secrets of fragmentary texts.” -  Heinz Faßmann, President, Austrian Academy of Sciences

The Austrian Academy of Sciences (OeAW) is pioneering the use of AI to revolutionize the study of antiquity. Partnering with Mistral, OeAW is developing Apollo, the world’s first advanced  Large Language Model for Ancient Greek. Trained on the largest digital corpus of historical Greek to date, Apollo helps researchers reconstruct damaged texts and accelerate discoveries that once took years into hours, with future phases planned to support semantic search and handwritten inscription decipherment. 

Decoding a millennium of history

For centuries, scholars have grappled with the limitations of manual transcription and interpretation of Ancient Greek texts. With over one million unread Greek papyri worldwide, including tens of thousands held by the Austrian National Library, the sheer volume of fragmentary inscriptions and documents has made comprehensive analysis a daunting task. Traditional methods are slow, labor-intensive, and often unable to reconstruct heavily damaged texts.

The OeAW’s Austrian Archaeological Institute sought a solution that could bridge the gap between cutting-edge technology and the nuanced demands of historical research. They needed an AI system capable of understanding a language that evolved over millennia, while adhering to the rigorous standards of academic scholarship. As Anna Dolganov, the project’s lead and a papyrologist at the OeAW, explains: “This LLM marks the beginning of an exciting journey in the study of antiquity. One where AI can finally help us unlock the knowledge trapped in these ancient fragments.”

Introducing Apollo, a custom LLM for Ancient Greek

Apollo is not just another language model. It is the first of its kind: an LLM trained on a specialized corpus of 600 million words from historical Greek texts, alongside tens of thousands of published inscriptions and papyri. This enables Apollo to tackle tasks previously thought impossible, such as reconstructing missing sections of texts or identifying thematic connections across vast collections.

Mistral’s expertise in building state-of-the-art models, combined with services partner Reply , provided the foundation for Apollo’s development. The collaboration ensured a secure, European-based infrastructure, allowing researchers to harness the full potential of generative AI while maintaining strict data protection and security standards. “The partnership with Mistral and Reply has given us the tools to turn a vision into reality; one that respects the complexity of Ancient Greek and the needs of modern scholarship,” says Dolganov.

A new era for humanities research

Apollo is already transforming how researchers interact with ancient texts. What once required years of painstaking work can now be achieved in hours, freeing scholars to focus on interpretation and analysis. The model’s ability to conduct semantic searches and restore damaged texts opens new avenues for understanding our shared history, from daily life in ancient societies to the evolution of language itself.

For OeAW, this project is more than a technical achievement, it’s a testament to how AI can serve as a force multiplier for the humanities. As President Faßmann notes, “This collaboration demonstrates how AI is advancing research in fields where precision and context are everything. Apollo is not just a tool; it’s a new way of seeing the past.”

With Apollo, the OeAW is not only preserving history but also redefining how we access it.