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

Can I run Qwen3.6-27B on an RTX 3090?

Yes, with a GGUF version

The published weights are too big, but the UD-Q5_K_XL file from unsloth/Qwen3.6-27B-GGUF (20.0 GB) fits, with room for ≈ 21K tokens of context.

Full analysis of Qwen3.6-27B Everything that runs on RTX 3090 Get the UD-Q5_K_XL file on Hugging Face
ollama run hf.co/unsloth/Qwen3.6-27B-GGUF:UD-Q5_K_XL

Files and context on RTX 3090

Each file of Qwen3.6-27B (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 3090 Context (16-bit)Context (8-bit)
As published55.6 GBDoesn't fit——
BF16unsloth/Qwen3.6-27B-GGUF53.8 GBDoesn't fit——
UD-Q8_K_XLunsloth/Qwen3.6-27B-GGUF35.3 GBDoesn't fit——
Q8_0unsloth/Qwen3.6-27B-GGUF28.6 GBDoesn't fit——
UD-Q6_K_XLunsloth/Qwen3.6-27B-GGUF25.6 GBDoesn't fit——
Q6_Kunsloth/Qwen3.6-27B-GGUF22.5 GBDoesn't fit——
UD-Q5_K_XLunsloth/Qwen3.6-27B-GGUF20.0 GBTight21K42K
Q5_K_Munsloth/Qwen3.6-27B-GGUF19.5 GBTight29K58K
Q5_K_Sunsloth/Qwen3.6-27B-GGUF19.0 GBTight37K74K
UD-Q4_K_XLunsloth/Qwen3.6-27B-GGUF17.6 GBTight57K114K
Q4_1unsloth/Qwen3.6-27B-GGUF17.3 GBFits62K125K
Q4_K_Munsloth/Qwen3.6-27B-GGUF16.8 GBFits69K138K
IQ4_NLunsloth/Qwen3.6-27B-GGUF16.1 GBFits80K160K
Q4_K_Sunsloth/Qwen3.6-27B-GGUF15.9 GBFits83K167K
Q4_0unsloth/Qwen3.6-27B-GGUF15.8 GBFits84K169K
IQ4_XSunsloth/Qwen3.6-27B-GGUF15.4 GBFits89K179K
UD-Q3_K_XLunsloth/Qwen3.6-27B-GGUF14.5 GBFits104K208K
Q3_K_Munsloth/Qwen3.6-27B-GGUF13.6 GBFits117K234K
Q3_K_Sunsloth/Qwen3.6-27B-GGUF12.4 GBFits135K256K
UD-IQ3_XXSunsloth/Qwen3.6-27B-GGUF12.0 GBFits141K256K
UD-Q2_K_XLunsloth/Qwen3.6-27B-GGUF11.8 GBFits143K256K
UD-IQ2_Munsloth/Qwen3.6-27B-GGUF10.8 GBFits158K256K
UD-IQ2_XXSunsloth/Qwen3.6-27B-GGUF9.39 GBFits180K256K

Qwen3.6-27B on other hardware

Popular models that run on RTX 3090

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.