AI Matters
Why it matters
NVIDIA releases open speech-to-speech model with ultra-low latency
NVIDIA released NemotronLabs VoiceChat 11B, an open-source, full-duplex speech-to-speech model featuring a 450-millisecond response latency.
Why it matters
Developers can build highly responsive, natural voice assistants that can call tools and respond to users in under half a second.
Latest brief · Aug 10, 2026