Nemotron-3-Nano-Omni-30B-A3B-Reasoning-FP8

Home/Models/nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-FP8

nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-FP8

kind: fp8
quant: fp8 (8-bit)
family: nemotron
modality: any-to-any
root model: nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16
provider: nvidia
license: other
language: —
context: 256k
library: transformers
vRAM (GB): —
size on disk: —
downloads: 868,610
likes: 60
created: —
last modified: 2026-05-05T14:35:05+00:00 (2026)
huggingface: https://huggingface.co/nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-FP8
tags: transformerssafetensorsNemotronH_Nano_Omni_Reasoning_V3feature-extractionnvidiapytorchmultimodalany-to-anycustom_codedataset:nvidia/Nemotron-Image-Training-v3arxiv:2604.24954base_model:nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16base_model:quantized:nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16license:othermodeloptdeploy:azureregion:usdeploy:sagemaker

benchmark	category	score	context	conf	evidence
MATH	math	82.8	Math Reasoning	medium	`\| Math Reasoning \| MathVista_MINI \| 82.8 \| 75.5 \| 8.82 \|`
MMAU	agentic	74.62	MMAU	medium	`\| MMAU \| 74.62 \| 74.56 \| 74.34 \|`
MathVista	multimodal	71.9	MathVista_MINI	medium	`\| MathVista_MINI \| 71.90 \| 71.05 \| 71.30 \|`
LongBench	long-context	46.1	MMlongBench Doc	medium	`\| MMlongBench Doc \| 46.10 \| 45.84 \| 45.78 \|`
WER	speech-asr	5.95	HF-ASR (WER↓)	medium	`\| HF-ASR (WER↓) \| 5.95 \| 5.97 \| 5.95 \|`
WER	speech-asr	3.11	Tedium Long (WER↓)	medium	`\| Tedium Long (WER↓) \| 3.11 \| 3.12 \| 3.04 \|`

none scraped.