frostedunicorn/logos-v61-sft

TEXT GENERATIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:7.6BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Sep 17, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

frostedunicorn/logos-v61-sft is a 7.6 billion parameter research-grade language model, fine-tuned from Qwen2.5-7B-Instruct, specializing in Abrahamic religious texts across Arabic, Hebrew, Greek, English, and Latin. It is designed for multilingual text understanding, verse recall, and cross-tradition comparison within religious contexts. The model is optimized for research into religious text processing and understanding, not for authoritative interpretation or advice.

Loading preview...

Logos v6.1-SFT: Abrahamic Religious Texts Model

Logos v6.1-SFT is a 7.6 billion parameter research-grade language model, fine-tuned from Qwen2.5-7B-Instruct, specifically designed for working with Abrahamic religious texts. It supports multiple languages including Arabic, Hebrew, Greek, English, and Latin, making it a valuable tool for multilingual religious studies.

Key Capabilities

  • Multilingual Text Understanding: Processes religious texts across five languages.
  • Instruction-Tuned Q&A: Trained on 30K Q&A samples for verse recall, cross-tradition comparison, and tafsir explanation.
  • Citation Formatting (Experimental): Includes training on Quran, Bible, and Talmud citation formats, though accuracy is currently low (20%).
  • Research Focus: Developed for academic and research purposes in religious text analysis.

Limitations and Considerations

It is crucial to note that Logos v6.1-SFT is a research alpha model with known limitations. It exhibits low citation accuracy (20%), zero exact verse recall (relies on RAG), and a 50% hallucination rate, making it unsuitable for production factual queries or authoritative religious advice. The model has not memorized verse text and does not provide scholarly-grade Arabic diacritics. Users should be aware of these constraints and use the model strictly for research and experimental purposes.