Model Anatomy See what's inside open-weight AI models

Can I run Huihui-DeepSeek-V4.1-Flash-1017-abliterated-GGUF on an RTX 5090?

Yes

Its Huihui-DeepSeek-V4.1-Flash-1017-patch-Q2-1 file (3.66 GB) fits in the 28.8 GB an RTX 5090 can use, with room for an unknown amount of context.

Full analysis of Huihui-DeepSeek-V4.1-Flash-1017-abliterated-GGUF Everything that runs on RTX 5090 Get the Huihui-DeepSeek-V4.1-Flash-1017-patch-Q2-1 file on Hugging Face
ollama run hf.co/huihui-ai/Huihui-DeepSeek-V4.1-Flash-1017-abliterated-GGUF

Files and context on RTX 5090

Each file of Huihui-DeepSeek-V4.1-Flash-1017-abliterated-GGUF (and of GGUF versions the site has analyzed), and how much context fits next to it, with the context cache in 16-bit or 8-bit.

FileSizeOn RTX 5090 Context (16-bit)Context (8-bit)
UD-IQ1_M250 GBDoesn't fit——
UD-IQ1_S233 GBDoesn't fit——
Huihui-DeepSeek-V4.1-Flash-1017-patch-Q2-2-344.7 GBDoesn't fit——
Huihui-DeepSeek-V4.1-Flash-1017-patch-Q2-2-241.7 GBDoesn't fit——
Huihui-DeepSeek-V4.1-Flash-1017-patch-Q2-2-136.9 GBDoesn't fit——
Huihui-DeepSeek-V4.1-Flash-1017-patch-Q2-2-434.2 GBDoesn't fit——
Huihui-DeepSeek-V4.1-Flash-1017-patch-Q2-2-528.7 GBDoesn't fit——
Huihui-DeepSeek-V4.1-Flash-1017-patch-Q2-2-026.4 GBDoesn't fit——
Huihui-DeepSeek-V4.1-Flash-1017-patch-Q2-13.66 GBFitsnot knownnot known
IQ1_S1.79 GBFitsnot knownnot known
Huihui-DeepSeek-V4.1-Flash-1017-patch-GSQ-RCO-Coder-IQ2_XXS1.48 GBFitsnot knownnot known
IQ2_XXS1.16 GBFitsnot knownnot known
UD-IQ3_S802 MBFitsnot knownnot known
UD-IQ3_XXS802 MBFitsnot knownnot known
UD-IQ2_XXS782 MBFitsnot knownnot known
UD-Q2_K_XL782 MBFitsnot knownnot known
DeepSeek-V4.1-Flash-1017-UD-IQ1_M-patch619 MBFitsnot knownnot known
DeepSeek-V4.1-Flash-1017-UD-IQ1_S-patch608 MBFitsnot knownnot known

Huihui-DeepSeek-V4.1-Flash-1017-abliterated-GGUF on other hardware

Popular models that run on RTX 5090

Assumes 90% of the card's memory is usable and at least 4K tokens of context for one conversation, with the context cache in 16-bit unless noted. Fits leaves at least 20% of that memory for context; Tight leaves less.