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

Can I run Qwen3.6-35B-A3B on an RTX 3090?

Yes, with a GGUF version

The published weights are too big, but the UD-Q4_K_S file from unsloth/Qwen3.6-35B-A3B-GGUF (20.9 GB) fits, with room for ≈ 31K tokens of context.

Full analysis of Qwen3.6-35B-A3B Everything that runs on RTX 3090 Get the UD-Q4_K_S file on Hugging Face
ollama run hf.co/unsloth/Qwen3.6-35B-A3B-GGUF:UD-Q4_K_S

Files and context on RTX 3090

Each file of Qwen3.6-35B-A3B (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 published71.9 GBDoesn't fit——
BF16unsloth/Qwen3.6-35B-A3B-GGUF69.4 GBDoesn't fit——
UD-Q8_K_XLunsloth/Qwen3.6-35B-A3B-GGUF38.5 GBDoesn't fit——
Q8_0unsloth/Qwen3.6-35B-A3B-GGUF36.9 GBDoesn't fit——
UD-Q6_K_XLunsloth/Qwen3.6-35B-A3B-GGUF31.8 GBDoesn't fit——
UD-Q6_Kunsloth/Qwen3.6-35B-A3B-GGUF29.3 GBDoesn't fit——
UD-Q5_K_XLunsloth/Qwen3.6-35B-A3B-GGUF26.6 GBDoesn't fit——
UD-Q5_K_Munsloth/Qwen3.6-35B-A3B-GGUF26.5 GBDoesn't fit——
UD-Q5_K_Sunsloth/Qwen3.6-35B-A3B-GGUF24.9 GBDoesn't fit——
UD-Q4_K_XLunsloth/Qwen3.6-35B-A3B-GGUF22.4 GBDoesn't fit——
UD-Q4_K_Munsloth/Qwen3.6-35B-A3B-GGUF22.1 GBDoesn't fit——
MXFP4_MOEunsloth/Qwen3.6-35B-A3B-GGUF21.7 GBDoesn't fit——
UD-Q4_K_Sunsloth/Qwen3.6-35B-A3B-GGUF20.9 GBTight31K61K
Qwen3.6-35B-A3B-UD-IQ4_NL_XLunsloth/Qwen3.6-35B-A3B-GGUF19.5 GBTight97K194K
UD-IQ4_NLunsloth/Qwen3.6-35B-A3B-GGUF18.0 GBTight167K256K
UD-IQ4_XSunsloth/Qwen3.6-35B-A3B-GGUF17.7 GBTight181K256K
UD-Q3_K_XLunsloth/Qwen3.6-35B-A3B-GGUF16.8 GBFits224K256K
UD-Q3_K_Munsloth/Qwen3.6-35B-A3B-GGUF16.6 GBFits235K256K
UD-Q3_K_Sunsloth/Qwen3.6-35B-A3B-GGUF15.4 GBFits256K256K
UD-IQ3_Sunsloth/Qwen3.6-35B-A3B-GGUF13.7 GBFits256K256K
UD-IQ3_XXSunsloth/Qwen3.6-35B-A3B-GGUF13.2 GBFits256K256K
UD-Q2_K_XLunsloth/Qwen3.6-35B-A3B-GGUF12.3 GBFits256K256K
UD-IQ2_Munsloth/Qwen3.6-35B-A3B-GGUF11.5 GBFits256K256K
UD-IQ2_XXSunsloth/Qwen3.6-35B-A3B-GGUF10.8 GBFits256K256K
UD-IQ1_Munsloth/Qwen3.6-35B-A3B-GGUF10.0 GBFits256K256K

Qwen3.6-35B-A3B 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.