NVIDIANemotron 3
Nemotron 3 Nano Omni 30B-A3B Reasoning
Nano Omni accepts video, audio, images and text, and returns text. Its card reports 31B total parameters and about 3B active per token.
What the release provides
The maximum context is 256K. Reasoning is enabled by default and can be changed with enable_thinking. Separate BF16, FP8 and NVFP4 weights have different documented hardware requirements.Nemotron 3 Nano Omni 30B-A3B Reasoning official model card
How to use it well
Practical guidance. For a meeting or training video, preserve timestamps and distinguish what was said from what appeared on screen. Ask for evidence from both channels when they conflict. A unified model can still miss a short visual event or quiet speech.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16 |