NemotronLabs has introduced VoiceChat, an open-weight full-duplex speech-to-speech model designed to integrate listening, transcription, reasoning, and speaking within a unified streaming architecture. The system combines a streaming speech encoder and decoder-only language model with parallel specialized output streams for agent text and structured function calls, alongside an auxiliary RNN-T branch for incremental user transcription.

  • On Full-Duplex-Bench 1.0, the model achieves the lowest pause-handling takeover rates among evaluated open-weight systems, with 100% takeover following user interruptions and a 4.33/5 post-interruption response-quality score.
  • It resumes responses after user backchannels in 93% of cases on Full-Duplex-Bench 1.5.
  • The model obtains a 55.1 normalized average on VoiceBench and achieves 82.5% tool-selection F1 on Full-Duplex-Bench 3.0, though argument accuracy and end-to-end tool execution require improvement.

These results demonstrate that full-duplex interaction, speech recognition and generation, general language capabilities, and external tool use can be integrated in a single open model without sacrificing real-time conversational behavior.