Hugging Face and Cerebras demo an open, modular speech-to-speech agent stack

The Neuron·low signal

Hugging Face and Cerebras showed an open, modular speech-to-speech stack combining Nvidia Parakeet for ASR, Google DeepMind's Gemma 4 31B running on Cerebras for reasoning, and Alibaba's Qwen3TTS for synthesis. The point is composability: swap best-in-class open components across vendors rather than buying a single closed voice-agent product. Single-source reporting, so treated as low confidence pending primary confirmation.