What changed
NVIDIA announced the release of Nemotron 3.5 Transcribe, an open-weight speech model that handles transcription, speaker attribution, and translation with a single prompt. The model is available on RTX Spark laptops and can process audio in 40 languages, including seven Indic languages. It ranks No. 1 on the Hugging Face Open ASR and No. 2 among open-weight models on Artificial Analysis, with approximately 4.3 billion active parameters.
What this means in practice
Developers can use Nemotron 3.5 Transcribe to transcribe recorded conversations, label speakers, and translate audio into English with plain-language prompts. The model supports specialized vocabulary in fields like healthcare, financial services, and customer service. It is available today on Hugging Face, with quantized versions coming soon for local deployment on NVIDIA RTX Spark and DGX Spark systems. Developers can also deploy the model using NeMo-Speech.cpp or vLLM.
Limitations and availability
The model is available on Hugging Face and will be deployable on NVIDIA RTX Spark and DGX Spark systems with quantized versions. The supplied announcement does not specify the exact release date for quantized versions. The model is part of NVIDIA's broader RTX Spark ecosystem, which includes new resource management features and support for Liquid AI's LFM3 family of models.
