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

Can I run Huihui-Qwen3.8-27B-abliterated-GGUF on an RTX 4090?

Yes, just

Its Q4_K_L file (20.9 GB) fits in the 21.6 GB an RTX 4090 can use, with room for ≈ 8K tokens of context.

Full analysis of Huihui-Qwen3.8-27B-abliterated-GGUF Everything that runs on RTX 4090 Get the Q4_K_L file on Hugging Face
ollama run hf.co/huihui-ai/Huihui-Qwen3.8-27B-abliterated-GGUF:Q4_K_L

Files and context on RTX 4090

Each file of Huihui-Qwen3.8-27B-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 4090 Context (16-bit)Context (8-bit)
BF1654.7 GBDoesn't fit——
Huihui-Qwen3.8-27B-abliterated-bf1654.7 GBDoesn't fit——
F1653.8 GBDoesn't fit——
Huihui-Qwen3.8-27B-abliterated-Q8_0_L38.8 GBDoesn't fit——
UD-Q8_K_XL31.5 GBDoesn't fit——
Q8_029.0 GBDoesn't fit——
Q8_K_L27.3 GBDoesn't fit——
Q6_K_L24.9 GBDoesn't fit——
UD-Q6_K_XL24.8 GBDoesn't fit——
Huihui-Qwen3.8-27B-abliterated-UD-DW-Q6_K_L23.9 GBDoesn't fit——
Q5_K_L22.9 GBDoesn't fit——
Q6_K22.4 GBDoesn't fit——
Q4_K_L20.9 GBTight8K16K
UD-Q5_K_XL20.7 GBTight12K23K
Q5_K19.5 GBTight28K57K
Q3_K_L18.7 GBTight41K82K
UD-Q4_K_XL17.4 GBTight61K121K
Q2_K_L17.2 GBFits63K126K
Q4_K16.8 GBFits69K138K
Q4_K_M16.6 GBFits73K146K
Q4_K_S15.6 GBFits87K174K
UD-IQ4_XS14.4 GBFits105K210K
Q3_K13.5 GBFits118K237K
UD-Q3_K_XL13.3 GBFits121K242K
Huihui-Qwen3.8-27B-abliterated-UD-DW-IQ3_S12.3 GBFits136K256K
Huihui-Qwen3.8-27B-abliterated-Swift-1.5-GSQ-RCO-IQ3_S-mtp12.1 GBFits139K256K
IQ3_S12.1 GBFits139K256K
UD-IQ3_S12.0 GBFits141K256K
Huihui-Qwen3.8-27B-abliterated-Swift-1.5-GSQ-RCO-IQ3_S11.8 GBFits144K256K
Huihui-Qwen3.8-27B-abliterated-GSQ-RCO-IQ3_S11.8 GBFits144K256K
Huihui-Qwen3.8-27B-abliterated-UD-DW-IQ3_XXS11.4 GBFits150K256K
UD-IQ3_XXS11.0 GBFits155K256K
Q2_K10.9 GBFits158K256K
Huihui-Qwen3.8-27B-abliterated-Swift-1.5-GSQ-RCO-IQ3_XXS-mtp10.4 GBFits164K256K
IQ3_XXS10.4 GBFits164K256K
Huihui-Qwen3.8-27B-abliterated-Swift-1.5-GSQ-RCO-IQ3_XXS10.1 GBFits169K256K
Huihui-Qwen3.8-27B-abliterated-GSQ-RCO-IQ3_XXS10.1 GBFits169K256K
UD-Q2_K_XL10.0 GBFits170K256K
IQ2_S9.61 GBFits176K256K
Huihui-Qwen3.8-27B-abliterated-Swift-1.5-GSQ-RCO-IQ2_S9.26 GBFits182K256K
UD-IQ2_S8.78 GBFits189K256K
IQ2_XS8.76 GBFits189K256K
Huihui-Qwen3.8-27B-abliterated-UD-IQ2_S8.42 GBFits194K256K
Huihui-Qwen3.8-27B-abliterated-Swift-1.5-GSQ-RCO-IQ2_XS8.42 GBFits194K256K
Huihui-Qwen3.8-27B-abliterated-Underdog-Saluki-Swift-1.5-IQ2-mtp8.25 GBFits197K256K
Huihui-Qwen3.8-27B-abliterated-Underdog-Saluki-Swift-1.5-IQ27.90 GBFits202K256K
Huihui-Qwen3.8-27B-abliterated-Ternary-Bonsai-PQ2_07.70 GBFits205K256K
Huihui-Qwen3.8-27B-abliterated-Ternary-Bonsai-PTQ1_06.61 GBFits221K256K

Huihui-Qwen3.8-27B-abliterated-GGUF on other hardware

Popular models that run on RTX 4090

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.