Austrian Academy of Science to release Apollo, first advanced Greek language model
The Austrian Academy of Science announced that on Wednesday it will release Apollo, the world’s first advanced large language model for Ancient Greek. Developed with French AI lab Mistral and technology services firm Sail Reply, the model was trained on roughly 600 million historical Greek words taken from manuscripts, papyri and inscriptions. Apollo will be offered for free to scholars through…
Key points
- Apollo was trained on about 600 million ancient Greek words from manuscripts, papyri and inscriptions.
- The model will be free for academics via a chatbot that proposes multiple word options for gaps.
- Developed by the Austrian Academy of Science with Mistral and Sail Reply, it adapts to dialects such as Doric.
Researchers say the tool could speed up the painstaking work of reconstructing fragmented documents. Professors Stephen Colvin (UCL) and Anna Dolganov (Austrian Academy) note that Apollo can adapt to dialects such as Doric and supplement Homeric Greek. While the model is not expected to overturn broad historical narratives, it may reveal new details about everyday life in antiquity. Critics warn that relying on probabilistic predictions could introduce errors, so Apollo presents multiple word options for human experts to choose from. If successful, the approach could be extended to other ancient languages like Latin or Egyptian.
Model page: Apollo →
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us.
More in Generative AI & Models
All →- Tencent launches Hy Image 3.5 preview, says it matches ByteDance's Seedream 5.0 Pro · 1 src
- Hugging Face adds GGUF support to Transformers for local inference · 1 src
- Rumors of new AI models may surface next week · 1 src
- Opinion: users complain about preachy ChatGPT, OpenAI tweaks GPT‑5.3 Instant · 1 src
- Xiaomi releases MiMo-V2.6-Pro-RL with 1M-token context and 1.02T parameters · 4 src
Comments
via GitHub Discussions