Nvidia Nemotron 3 Nano 30B A3B BF16
NVIDIA develops the Nvidia Nemotron 3 Nano 30B A 3B FP 8, a large language model designed for unified reasoning and non-reasoning tasks, exceling in math-related tasks as evidenced by its top 25% Math Index score. It features a hybrid Mixture-of-Experts architecture with 23 MoE layers and 6 Attention layers, allowing for efficient processing of long context windows up to 262,144 tokens.
Input
Output
Context
262K
Max Output
20K
Parameters
31.6B
Technical Specifications
Model TypeChat
Context Window262,144 tokens
Max Output Tokens20,480 tokens
Parameters31.6B
Release DateDec 4, 2025
Training CutoffNov 28, 2025
Licenseother
Open SourceYes
Input Modalities
Text
Output Modalities
Text
Capabilities
Benchmarks
Artificial Analysis
10.2%
HLE
75.7%
GPQA
29.6%
SciCode
79.4%
MMLU Pro
91.0
Math
19.0
Coding
74.1%
LiveCodeBench
17.5
Intelligence
51.8
Speed (tok/s)
42.9
TTFA (s)
4.2
TTFT (s)
Resources & Links
Estimated GPU Requirements for Nvidia Nemotron 3 Nano 30B A3B BF16
Estimates based on INT8 quantization. Actual requirements vary by framework and configuration.
Related Tools
Data sourced from official provider APIs and documentation
Last updated: Jun 24, 2026
Start building with the right model.
Automatically route workloads to the right model for every task, every time.