A developer has released a complete GGUF package for NVIDIA’s Nemotron-3-Nano-Omni-30B model, adding previously missing audio and video support to the llama.cpp ecosystem. The release includes nine quantizations ranging from Q3_K_M to BF16 and a unified projector file that integrates both C-RADIO vision and Parakeet audio encoders.

  • A dedicated llama.cpp fork implements the Parakeet/FastConformer audio graph and C-RADIO video graph, enabling one-pass processing of video with its soundtrack.
  • The video graph matches NVIDIA’s PyTorch reference to 0.0019 percent relative L2 error, while the audio graph transcribes test narration word for word.
  • Prebuilt Linux arm64 CUDA 13 binaries are provided for hardware such as DGX Spark and Jetson Thor.

This update allows users to run image, text, audio, and video modalities simultaneously through a single llama-server instance on compatible hardware.