NVIDIANemotron 3
Nemotron 3 Nano 30B-A3B
Nemotron 3 Nano 30B is a text model with hybrid Mamba-2, expert and attention layers. The card reports 3.5B active parameters.
What the release provides
Each expert layer includes 128 routed experts plus a shared expert, with six routed experts active per token. The published input limit is one million tokens.Nemotron 3 Nano 30B-A3B official model card
How to use it well
Practical guidance. Treat the A3B family name as a model identifier rather than an exact runtime-memory figure. Measure your chosen precision at the expected concurrency. A smaller active parameter count does not remove the memory cost of the other experts.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16 |
| Release license | NVIDIA Nemotron Open Model LicenseNemotron 3 Nano 30B-A3B official model card |