NVIDIA Nemotron Nano 2: An Accurate and Efficient Hybrid Mamba-Transformer Reasoning Model
- Type
- paper
- Venue
- arXiv
- Year
- 2026
- Source
- huggingface
- Access
- public
- Language
- en
- Added
- 2026-09-29
- Verified
- 2026-09-29
Summary
Technical report for NVIDIA's Nemotron Nano 2 family (9B/12B hybrid Mamba-Transformer reasoning models, 128K context) trained on the Nemotron pretraining dataset that includes Nemotron-CC-v2.
Research notes
- Key findings: Hybrid Mamba-Transformer reasoning models at 9B/12B scale with 128K context, trained on the Nemotron pretraining dataset
- Cited on the Nemotron-CC-v2 dataset card as the model family the data supports.