Skip to main content
Language model

Nemotron 3 Nano 30B A3B

NVIDIA

Released
December 15, 2025
Data date
September 26, 2026

Nemotron 3 Nano 30B A3B is the specific BF16 text checkpoint in the Nemotron 3 Nano line. It has 30 billion total parameters, activates 3.5 billion per token, and combines the hybrid architecture with reasoning and tool use. The identifier therefore names a particular weight and configuration form, not only a product family.

The model card documents separate thinking and direct-response modes with a configurable thinking budget. NVIDIA specifies up to 1M context tokens, while the packaged configuration defaults to 262,144 tokens.

Specifications and access

SpecificationValue and source
Weights
Open weightsSource
Access
Parameters
30 billionSource
License
NVIDIA Open Model LicenseSource
Inputs
TextSource
Outputs
TextSource
Reasoning
Tool calling
Active parameters
3.5 billionSource
Maximum context window
Up to 1M tokensSource
Maximum output
1M tokensSource
Packaged default configuration
262,144 tokensSource