Back to calculator

Model notes

OpenReasoning Nemotron 14B

Mid-sized dense Nemotron checkpoint for users who want stronger reasoning behavior than 7B without stepping straight into 32B deployment territory.

14.7B dense • 131,072 context • 8 KV heads

Architecture

Model spec

Architecture

Dense decoder-only transformer

Total params

14.7B

Active params

Dense model

Layers

48

Hidden size

5,120

Attention heads

40

KV heads

8

KV-bearing layers

48

Context length

131,072

Modality

Text

License

CC-BY-4.0 + Apache 2.0

Why it matters

Why memory behaves this way

Research highlight

The reasoning post-training is layered on top of the Qwen2.5 14B architecture, so the model keeps its dense long-context grouped-attention profile.

Memory note

This is still a dense 14B-class checkpoint: weights dominate the fit decision, and context length becomes the next major lever after quantization.

Checkpoints

Official profiles

Official BF16 checkpoint

BF16 checkpoint

Current

NVIDIA publishes OpenReasoning-Nemotron-14B as a Transformers-format dense derivative of Qwen2.5-14B, and v1 models it accordingly across the runtime presets.

vLLMTransformers
Open checkpoint

Sources

Reference links