Skip to content

Nemotron 3 Nano Omni

Released 2026-04 · multimodal · 256K tokens · 0 benchmarks

Editorial notes

Lanzado 28 abril 2026. Omni multimodal: text+image+video+audio in, text out. 30B total / 3B active, 128 experts MoE + 1 shared, hybrid Mamba-Transformer. Vision encoder C-RADIOv4-H, audio Parakeet-TDT-0.6B-v2. Lider en MMlongbench-Doc, OCRBenchV2 English, WorldSense, DailyOmni, VoiceBench. 9x throughput vs alternatives en MediaPerf. NVIDIA Open Model License (uso comercial permitido). Free en OpenRouter.

Spec sheet

Company
Nvidia
Country
US
Type
multimodal
Release
2026-04
Context
256K tokens
Slug
nemotron-3-nano-omni
No scores published for this model. The editorial notes have additional context.
Cite this model
BibTeX · APA

BibTeX

@misc{frontier-nemotron-3-nano-omni,
  title  = {Nemotron 3 Nano Omni},
  author = {{Nvidia}},
  year   = {2026},
  note   = {Frontier Benchmarks AI atlas. Accessed 2026-05-08},
  url    = {https://frontierbenchmarks.com/models/nemotron-3-nano-omni}
}

APA

Nvidia (2026). Nemotron 3 Nano Omni [Large language model]. Frontier Benchmarks AI. Retrieved 2026-05-08, from https://frontierbenchmarks.com/models/nemotron-3-nano-omni

Citation reflects the atlas page, not the original model paper. For the paper, see the "Resources" section above.