Austrian Academy of Sciences is developing the Ancient Greek AI “Apollo” with Mistral AI and Reply

Jump to

Header (Main-Nav & Top-Nav)

Content

Footer

JavaScript is needed to use this website in its full extent. Please activate JavaScript in your browser.

Back<br>InnovationsAncient studies<br>Austrian Academy of Sciences is developing the Ancient Greek AI “Apollo” with Mistral AI and Reply<br>Austrian Archaeological Institute Develops Custom AI Model for Hundreds of Thousands of Ancient Papyri and Inscriptions ++ Faßmann: “AI and ancient languages are not a contradiction”

18.03.2026

Open dialogue for sharing this pageShareShare via email<br>Share on LinkedIn<br>Share on Facebook<br>Share on Bluesky

The Austrian Academy of Sciences (OeAW) is collaborating with Mistral AI and Sail Reply, a Reply Group Company, on the development of a Large Language Model (LLM) for Ancient Greek: Apollo, named after the Greek god of light and patron of the arts and sciences, will propel research on ancient Greek texts. The model supports advanced searching and automatic text restoration in hundreds of thousands of undeciphered papyri and inscriptions, making it possible to accurately capture content in a matter of hours rather than years. The OeAW and its partners are doing pioneering work, as LLMs have not yet been developed for a historical language evolving over many centuries or the reconstruction of heavily damaged ancient texts.<br>Ancient language, new technology<br>On behalf of the OeAW, the project is led by Anna Dolganov, an ancient historian and papyrologist at the Austrian Archaeological Institute of the OeAW, who provides field–specific guidance, oversees the integration of relevant sources, and guarantees scientific quality. Through her expertise, Dolganov ensures that historical contextualization and methodological standards are upheld.<br>OeAW President Heinz Faßmann: “Ancient languages and artificial intelligence are not a contradiction. The collaboration between our Archaeological Institute and Mistral and Reply demonstrates how AI is advancing research in the Humanities. Apollo will support researchers toward a better understanding of often fragmentary ancient texts and thereby amplify our understanding of our shared history.”<br>1 million unread papyri<br>Anna Dolganov: “Our project with Mistral AI and Sail Reply is building the world’s first advanced multimodal Large Language Model for an ancient language, trained on the largest digital corpus of historical Greek to date. This AI system can be developed in many directions for a wide range of research tasks, from reconstructing fragmentary inscriptions and papyri to conducting semantic and thematic searches across the entire Greek textual tradition to deciphering handwritten texts. For example: there are one million Greek papyri worldwide that have never been read, tens of thousands of which are held by the Papyrus Collection of the Austrian National Library. Such treasures of historical knowledge are our target. This LLM marks the beginning of an exciting journey in the study of antiquity.”<br>Apollo is trained on a highly specialized corpus comprising ca. 600 million words from historical Greek texts and tens of thousands of published inscriptions and papyri. The collaboration with Mistal and Sail Reply provides a secure environment on European infrastructures and enables the full potential of generative AI to be harnessed while adhering to strict data protection and data security requirements.

Anna Dolganov

Anna Dolganov

No sudden upheaval: Researchers retell the long history of the Stone Age

No sudden upheaval: Researchers retell the long history of the Stone Age

Searchbar

Start search

Notice about cookies

This site uses cookies. They ensure essential functionalities of the website and enable us to continuously optimize content. Help us by agreeing to the collection of statistical data and the presentation of external multimedia content. You can change your cookie settings at any time. Further information can be found in the privacy settings and by clicking the Data Protection Link.

Settings

NecessaryExplanation for cookie group "Necessary"

These cookies are required for the smooth functioning of our website.

Name<br>Purpose<br>Storage duration<br>Type<br>Provider

CookieConsent

Remembers your consent to the use of cookies.

year

HTML

Web Consent

fe_typo_user

Assigns your browser to a session on the server. This only affects the content you see and is not evaluated or processed by us

HTTP

Web User

StatisticsExplanation for cookie group "Statistics"

With the help of these cookies, we strive to make our offer even more attractive for you. The user flow can be analyzed and assessed using pseudonymised data from website users. This gives us the opportunity to optimize website content.

Name<br>Purpose<br>Storage duration<br>Type<br>Provider

_pk_id

Used to store a few details about the user like unique visitor ID.

13<br>months

HTML

Matomo-id

_pk_ref

Used to store information about...