Dev Radar
Support
LiveUpdated 2026-11-03 23:16 UTC

deepseek-v4.1-flash-nvfp4

deepseek-v4.1-flash-nvfp4 — Image-Text-to-Text model with 763B parameters. Download, explore files and run it on…

deepseek-v4.1-flash-nvfp4 is Image-Text-to-Text model with 763B parameters. Download, explore files and run it on Hugging Face.. It is ranked #449 on the Dev Radar, in Other, first seen 47 days ago and shared in 5 posts (73.6k views).

Visit huggingface.co

nvidia/DeepSeek-V4.1-Flash-NVFP4 · Hugging Face Hugging Face Log In Sign Up…

What people said about deepseek-v4.1-flash-nvfp4 on X

🤯炸了!DeepSeek V4.1 Flash 1017 Obliterated 无审查魔改版来了!拒绝率从 81.62% 暴降至 4.30%,MMLU 还微升,能力几乎零损耗! 这是目前最硬核的本地魔改之一,用 OBLITERATUS 的 failspy 方法做拒绝方向消融,直接把“I can’t help with that”的习惯从权重里拔掉。 关键亮点: 🔹拒绝率:81.62% → 4.30%(-77.33 个百分点),50 条测试只拒了 2 条 🔹有害协助率:10.38% → 78.28%(模型更“听话”) 🔹能力保持:连贯性 100%,能力检查 6/6 全过 🔹MMLU:88.08% vs 原版…

— @NFT_Chen, 2 days ago · 405 likes · see the post

😳什么?!Terminal-Bench狂飙+11.1分!英伟达NVFP4魔改DeepSeek-V4.1-Flash-1017 杀到,Blackwell专属加速,性能成本直接起飞! 英伟达用Model Optimizer做W4A4 NVFP4量化,专打Blackwell FP4 Tensor Core,精度几乎无损,部署更轻、推理更快,适合Agent/RAG/商业私有化。 核心亮点 🔹基准全面提升,Agent与编码任务大幅领先上版 🔹NVFP4近无损量化,Blackwell硬件级加速 🔹1M上下文+多模态+MIT开源 🔹输出更短、吞吐更高、成本更低 快速上手…

— @NFT_Chen, 3 days ago · 73 likes · see the post

Alternatives to deepseek-v4.1-flash-nvfp4

deepseek-v4.1-flash-nvfp4 in numbers

FAQ

What is deepseek-v4.1-flash-nvfp4?

Image-Text-to-Text model with 763B parameters. Download, explore files and run it on Hugging Face. It was first shared on X 47 days ago and is ranked #449 on the Dev Radar.

Is deepseek-v4.1-flash-nvfp4 free?

Pricing is not stated on the page we read.

Who shared deepseek-v4.1-flash-nvfp4?

5 accounts on X, including @NFT_Chen, in 5 posts totalling 73.6k views.

Every post is read and classified by AGK Intelligence: what it is, which market it belongs to, and whether the link is a real tool. 31.3k posts from 5k X accounts over the last 21 days, 9.2k tools, 12 markets. New posts are collected several times an hour and the full ranking is recomputed every hour — last update 2026-11-03 23:16 UTC. Full method.