Model reference · open weights

personaplex

personaplex is an open-weight audio or speech model from NVIDIA. personaplex-7b-v1 (BF16) weighs 17.1 GB; the smallest configuration that runs it is RTX 4090 24 GB.

• PersonaPlex is a real-time speech-to-speech conversational model developed by NVIDIA that jointly performs streaming speech understanding and generation.
• It operates on continuous audio to support natural conversational dynamics, including interruptions and rapid turn-taking, while generating English speech responses.
• The model contains 8.4B parameters, uses a Transformer architecture based on Moshi, and is governed by the NVIDIA Open Model License Agreement.

Summary of the nvidia/personaplex-7b-v1 model card, 2026-10-01

Hugging Face ↗GitHub ↗Paper ↗