N
NVIDIA: Nemotron 3 Nano 30B A3B
◆ Strong7.6
/ 10
NVIDIA Nemotron 3 Nano 30B A3B is a small language MoE model with highest compute efficiency and accuracy for developers to build specialized agentic AI systems. The model is fully...
Specifications
| Attribute | Value |
|---|---|
| Lab | Nvidia |
| Tags | Fast Agentic |
| Overall Score | 7.6/10 |
| Release Date | 2025-12 |
| Context Window | 262,144 tokens |
| Input Price / 1M | $0.05 |
| Output Price / 1M | $0.20 |
| Input Modalities | Text |
| Output Modalities | Text |
Strengths
- Efficient MoE with 30B total / 3B active params
- Very low latency and cost
- Strong for agentic and tool-use tasks
- NVIDIA ecosystem integration
Weaknesses
- Limited deep reasoning capability
- Below larger Nemotron models
- Narrower general knowledge
Best For
- Latency-sensitive agent tasks
- Cost-efficient deployments
- NVIDIA-optimized inference
Sources & Further Reading
Related Models
Scores are aggregated from public benchmarks (MMLU, HumanEval, MATH, GSM8K, LMSYS) and normalized to a 1–10 scale. Methodology →