Qwen3-8B-DMS-8x

Home/Models/nvidia/Qwen3-8B-DMS-8x

nvidia/Qwen3-8B-DMS-8x

kind: original
quant: bf16 (16-bit)
family: qwen
modality: —
root model: Qwen/Qwen3-8B
provider: nvidia
license: other
language: —
context: 32,768,
library: transformers
vRAM (GB): —
size on disk: —
downloads: 238
likes: 37
created: —
last modified: 2026-01-22T16:07:37+00:00 (2026)
huggingface: https://huggingface.co/nvidia/Qwen3-8B-DMS-8x
tags: transformerssafetensorsqwen3nvidiapytorchkvcachecustom_codedataset:open-r1/OpenR1-Math-220karxiv:2506.05345arxiv:2505.09388arxiv:2305.20050arxiv:2107.03374arxiv:2311.07911base_model:Qwen/Qwen3-8Bbase_model:finetune:Qwen/Qwen3-8Blicense:othertext-generation-inferenceendpoints_compatibleregion:us

benchmark	category	score	context	conf	evidence
MATH-500	math	95.1	MATH-500	medium	`\| MATH-500 \| y \| 95.1 \| 95.5 \|`
IFEval	instruction-following	90.3	IFEval	medium	`\| IFEval \| y \| 90.3 \| 88.8 \|`
Chatbot Arena	chat	88.4	ArenaHard v0.1	medium	`\| ArenaHard v0.1 \| y \| 88.4 \| 89.7 \|`
HumanEval	coding	87.8	HumanEval	medium	`\| HumanEval \| y \| 87.8 \| 89.6 \|`
AIME 2024	math	75.0	AIME 2024	medium	`\| AIME 2024 \| y \| 75.0 \| 73.0 \|`
MMLU-Pro	knowledge	74.2	MMLU-Pro	medium	`\| MMLU-Pro \| y \| 74.2 \| 73.5 \|`
RULER	long-context	69.2	RULER 64K	medium	`\| RULER 64K \| n \| 69.2 \| 76.2 \|`
RULER	long-context	64.0		medium	`RULER 64`
GPQA Diamond	reasoning	58.8	GPQA Diamond	medium	`\| GPQA Diamond \| y \| 58.8 \| 57.6 \|`
RULER	long-context	25.0	RULER 128K	medium	`\| RULER 128K \| n \| 25.0 \| 21.4 \|`

none scraped.