Muslim: A Deployed Arabic Voice AI Platform for Grounded Islamic Knowledge
Muslim is a production Arabic voice AI for sourced Islamic knowledge, releasing a 6B router and TTS model.
Muslim is a production Arabic voice platform for sourced Islamic knowledge, combining NeMo Arabic ASR, an OpenAI-compatible LLM endpoint, self-hosted TTS, and retrieval across six Model Context Protocol servers. Released artifacts include Muslim-6B-PRO, a 5.94B tool-routing model, and Fasih-TTS-V1, ranked 5th of 17 overall and 2nd of 11 open-weight systems on the community Arabic TTS Arena for Modern Standard Arabic. The authors report 98.4% recitation-validation accuracy on 124 cases and 0.9–1.7 second end-to-end voice latency, plus account metering and observability aimed at a silent GPU-bound agent host.
- Voice stack uses NeMo Arabic ASR, an LLM endpoint, and self-hosted TTS.
- Muslim-6B-PRO is a 5.94B model for tool routing.
- Fasih-TTS-V1 ranks 5th of 17 and 2nd among 11 open-weight MSA systems.
- Recitation validation reached 98.4% accuracy on 124 cases.
- Measured end-to-end voice latency was 0.9 to 1.7 seconds.
- Vendors
- NVIDIA
- AI models
- Muslim-6B-PROFasih-TTS-V1
From the original article
1 min read at arxiv.orgWe present Muslim, a production Arabic voice AI platform serving grounded, sourced Islamic knowledge to real users. Beyond a real-time voice pipeline (NeMo Arabic ASR, an OpenAI-compatible LLM endpoint, self-hosted TTS) and a deterministic multi-source retrieval layer routed across six Model Context Protocol servers, we report three things a research prototype typically lacks.
By Yahya Mohamed Elnawasany · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.