Atlas

Models

← All models

Nemotron Nano 12B V2 VL BF16

NVIDIA · 2025-10-28 · 13.2B parameters

Nemotron Nano 12B V2 VL BF16 is the BF16 reference checkpoint of NVIDIA's Nemotron Nano V2 VL vision-language model, released on October 28, 2025; NVIDIA also ships FP8 (post-training-quantized) and NVFP4 (quantization-aware-distilled) variants. It combines the Nemotron Nano V2 12B hybrid Mamba-2/attention language backbone with a c-RADIOv2-H vision encoder, accepts text, multiple images and video, switches reasoning on or off, and supports a 131,072-token input-plus-output context. The checkpoint contains 13,181,860,358 parameters including the vision encoder and projector.

Benchmark scores

BenchmarkScore
AIME 202571.3
BLINK57.6
Capability127.2
ChartQA (test)89.8
CharXiv Reasoning41.7
CharXiv-D77.2
DocVQA94.7
GPQA Diamond64.1
IFEval78.2
IFEval (instruction-level strict accuracy)84.7
LiveCodeBench - July 2024 to January 202569.4
LongVideoBench - release-reported accuracy63.6
MATH-50096.9
MATH-Vision53.6
MathVista (mini)76.9
MMBench v1.1 (English, split unstated)83.0
MMLongBench-Doc32.1
MMMU68.0
MMStar71.7
OCRBench856
RealWorldQA - release-reported visual perception76.2
ScreenSpot-Pro5.5
ScreenSpot-v242.8
TextVQA (validation)85.4
Video-MME66.0
Loading Atlas data…