semiconductor / news / / TNW
Nemotron 3 Nano Omni unifies vision, audio, and language in a single architecture with 30B parameters but only 3B active per inference.
Nvidia's Nemotron 3 Nano Omni unifies vision, audio, and language into a single 30B-parameter model.
KEY POINTS
- The model uses only 3B active parameters per inference, enabling single GPU edge deployment.
- Nemotron 3 Nano Omni claims 9x throughput and 2.9x faster reasoning than comparable open multimodal models.
- It is the first open-weight multimodal model combining mixture-of-experts, unified architecture, and audio capabilities.
- Available under Nvidia's Open Model Agreement, it allows full commercial use and runs on multiple frameworks.
COMPANIES
Summarized by Newsio from TNW. How we summarize →