What are the hardware requirements for Qwen3.8 Flash Reasoning NVFP4 One Spark?

Qwen3.8 Flash Reasoning NVFP4 One Spark requires approximately 78.7 GB of VRAM. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Qwen3.8 Flash Reasoning NVFP4 One Spark and how large is it?

Qwen3.8 Flash Reasoning NVFP4 One Spark is developed by elisepaul, a model with 131.7b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

Where can I download or evaluate Qwen3.8 Flash Reasoning NVFP4 One Spark?

Qwen3.8 Flash Reasoning NVFP4 One Spark is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

Qwen3.8 Flash Reasoning NVFP4 One Spark by elisepaul — VRAM 78.7GB

Name: Qwen3.8 Flash Reasoning NVFP4 One Spark
Author: elisepaul

Qwen3.8 Flash Reasoning NVFP4 One Spark is an open-source language model by elisepaul. Features: 131.7b LLM, VRAM: 78.7GB, License: other, LLM Explorer Score: 0.3.

Arxiv:2307.08691 Arxiv:2312.06635 Arxiv:2403.13257 Arxiv:2404.19737 Arxiv:2409.19606 Arxiv:2412.15115 Arxiv:2412.19437 8-bit Agent Base model:deepseek-ai/deepsee... Base model:quantized:deepseek-... Blackwell Chain-of-thought Code Code-generation Coder Coding Conversational Cot Dataset:princeton-nlp/swe-benc... Deep-thinking Deepseek Deepseek-v4 Deepseek-v4.1-flash Deepseek-v4.1-flash-1017 Dgx Dgx-spark En Flashinfer Fp4 Fp8 Frankenmoe Gb10 Lpddr5x Mergekit Mixture-of-experts Model-merging Modelopt Moe Mtp Multi-token-prediction Mxfp8 Nvfp4 Nvidia Nvidia-modelopt One-spark Qwen Qwen3 Qwen3.8 Qwen4 exp Reasoning Region:us Safetensors Sharded Single-node Spark Speculative-decoding Swe-bench Tensorflow Unified-memory Vllm Zh

Model Card on HF 🤗: https://huggingface.co/elisepaul/Qwen3.8-Flash-Reasoning-NVFP4-One-Spark

Qwen3.8 Flash Reasoning NVFP4 One Spark Parameters and Internals

LLM Name	Qwen3.8 Flash Reasoning NVFP4 One Spark
Repository 🤗	https://huggingface.co/elisepaul/Qwen3.8-Flash-Reasoning-NVFP4-One-Spark
Base Model(s)	nvidia/Qwen3.8-Flash-Next-NVFP4 nvidia/DeepSeek-V4.1-Flash-1017-NVFP4 deepseek-ai/DeepSeek-V4.1-Flash nvidia/Qwen3.8-Flash-Next-NVFP4 nvidia/DeepSeek-V4.1-Flash-1017-NVFP4 deepseek-ai/DeepSeek-V4.1-Flash
Model Size	131.7b
Required VRAM	78.7 GB
Updated	2026-11-13
Maintainer	elisepaul
Model Type	qwen4_exp
Model Files	3.1 GB: 1-of-10 9.9 GB: 2-of-10 10.0 GB: 3-of-10 10.0 GB: 4-of-10 10.0 GB: 5-of-10 10.0 GB: 6-of-10 10.0 GB: 7-of-10 10.0 GB: 8-of-10 5.7 GB: 9-of-10 1.6 GB 3.0 GB
Supported Languages	en zh code
Model Architecture	Qwen4ExpForConditionalGeneration
License	other
Model Max Length	262144
Transformers Version	5.16.0.dev0
Tokenizer Class	Qwen2Tokenizer
Padding Token	<\|endoftext\|>
Errors	replace

Rank the Qwen3.8 Flash Reasoning NVFP4 One Spark capabilities

Have you tried this model? Rate its performance — this feedback helps the ML community find the right model for their needs.

Instruction Following and Task Automation
Factuality and Completeness of Knowledge
Censorship and Alignment
Data Analysis and Insight Generation
Text Generation
Text Summarization and Feature Extraction
Code Generation
Multi-Language Support and Translation