NVIDIA-Nemotron-3-Nano-4B-FP8

Home/Models/nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8

nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8

kind: fp8
quant: fp8 (8-bit)
family: nemotron
modality: text-generation
root model: nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16
provider: nvidia
license: other
language: en
context: 262K
library: transformers
vRAM (GB): —
size on disk: —
downloads: 91,506
likes: 28
created: —
last modified: 2026-03-20T20:50:46+00:00 (2026)
huggingface: https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8
tags: transformerssafetensorsnemotron_htext-generationnvidiapytorchconversationalcustom_codeendataset:nvidia/Nemotron-CC-v2dataset:nvidia/Nemotron-Post-Training-Dataset-v2dataset:nvidia/Nemotron-Science-v1dataset:nvidia/Nemotron-Instruction-Following-Chat-v1dataset:nvidia/Nemotron-Agentic-v1dataset:nvidia/Nemotron-Competitive-Programming-v1dataset:nvidia/Nemotron-Math-Proofs-v1dataset:nvidia/Nemotron-RL-Agentic-Conversational-Tool-Use-Pivot-v1dataset:nvidia/Nemotron-RL-instruction_followingdataset:nvidia/Nemotron-RL-agent-calendar_schedulingdataset:nvidia/Nemotron-RL-instruction_following-structured_outputsarxiv:2511.16664arxiv:2504.03624arxiv:2512.20856arxiv:2512.20848

benchmark	category	score	context	conf	evidence
MATH-500	math	95.4	MATH500	medium	`\| MATH500 \| 95.4 \| 95.5 \|`
IFEval	instruction-following	92.0	IFEVAL-Instruction	medium	`\| IFEVAL-Instruction \| 92 \| 91.7 \|`
RULER	long-context	91.1	RULER (128k)	medium	`\| RULER (128k) \| 91.1 \| 91.0 \|`
IFEval	instruction-following	88.0	IFEval-Instruction	medium	`\| IFEval-Instruction \| 88.0 \| 87.5 \|`
IFEval	instruction-following	87.9	IFEVAL-Prompt	medium	`\| IFEVAL-Prompt \| 87.9 \| 88.8 \|`
IFEval	instruction-following	82.8	IFEval-Prompt	medium	`\| IFEval-Prompt \| 82.8 \| 85.8 \|`
AIME 2025	math	78.5	AIME25	medium	`\| AIME25 \| 78.5 \| 76.3 \|`
BFCL (function calling)	agentic	61.1	BFCL v3	medium	`\| BFCL v3 \| 61.1 \| 60.6 \|`
GPQA	reasoning	53.2	GPQA-D	medium	`\| GPQA-D \| 53.2 \| 54.6 \|`
LiveCodeBench	coding	51.8	LiveCodeBench v5	medium	`\| LiveCodeBench v5 \| 51.8 \| 52.9 \|`
IFBench	instruction-following	44.2	IFBench-Instruction	medium	`\| IFBench-Instruction \| 44.2 \| 44.8 \|`
IFBench	instruction-following	43.2	IFBench-Prompt	medium	`\| IFBench-Prompt \| 43.2 \| 43.9 \|`

none scraped.