Nemotron-H-8B-Reasoning-128K

Home/Models/nvidia/Nemotron-H-8B-Reasoning-128K

nvidia/Nemotron-H-8B-Reasoning-128K

kind: original
quant: bf16 (16-bit)
family: nemotron
modality: text-generation
root model: nvidia/Nemotron-H-8B-Reasoning-128K
provider: nvidia
license: other
language: en
context: 128K
library: transformers
vRAM (GB): —
size on disk: —
downloads: 2,020
likes: 28
created: —
last modified: 2025-07-11T16:34:34+00:00 (2025)
huggingface: https://huggingface.co/nvidia/Nemotron-H-8B-Reasoning-128K
tags: transformerssafetensorsnvidiapytorchtext-generationconversationalenarxiv:2504.03624arxiv:2505.00949license:otherendpoints_compatibleregion:us

benchmark	category	score	context	conf	evidence
MATH-500	math	94.8	MATH500	medium	`\| MATH500 \| 94.8% \| 94.0% \| 94.0% \|`
MBPP	coding	84.1	MBPP	medium	`\| MBPP \| 84.1% \| 86.0% \| 86.2% \|`
IFEval	instruction-following	79.3	IFEVAL-Ins.	medium	`\| IFEVAL-Ins. \| 79.3% \| 79.6% \| 79.9% \|`
IFEval	instruction-following	71.8	IFEVAL-Prompt	medium	`\| IFEVAL-Prompt \| 71.8% \| 71.4% \| 71.5% \|`
MBPP	coding	71.4	MBPP\_PLUS	medium	`\| MBPP\_PLUS \| 71.4% \| 74.1% \| 73.3% \|`
BFCL (function calling)	agentic	63.6	BFCL	medium	`\| BFCL \| 63.6% \| 68.8% \| 70.6% \|`
GPQA	reasoning	53.5	GPQA	medium	`\| GPQA \| 53.5% \| 55.1% \| 54.5% \|`
AIME 2025	math	51.4	AIME25	medium	`\| AIME25 \| 51.4% \| 51.7% \| 46.7% \|`

none scraped.