Adventists AI has released Sophon-Realtime v1, an open-source full-duplex voice model family. This model family excels in understanding conversational nuances like tone, hesitation, and laughter, and can respond with similar paralinguistic cues. It achieves strong performance on benchmarks such as VoiceBench and Big Bench Audio, outperforming other open full-duplex models in paralinguistic understanding. The model is designed for efficient deployment, with different builds optimized for voice quality, user capacity, or speed, and is available under the Apache-2.0 license for commercial use. AI
IMPACT This release offers a powerful open-source alternative for real-time conversational AI, potentially accelerating development in voice assistants and interactive agents.
RANK_REASON Open-source release of a new voice model with advanced full-duplex and paralinguistic capabilities. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
Read on Hugging Face Trending Models →
- Adventists AI
- Apache Software License 2.0
- Big Bench Audio
- DeepSeek V4 Pro
- Gemini 3.8 Flash
- GPT-4o-mini-Audio
- Grok
- OpenAI GPT-Live
- Qwen3-Omni
- Sophon-Realtime-v1
- tensorrt
- vllm-omni
- VoiceBench
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →