GGPUfit

unsloth

Qwen3 Coder 480B A35B Instruct

Qwen3 Coder 480B A35B Instruct is a 480.2B mixture-of-experts qwen3moe model from unsloth, with about 35.0B parameters active per token. Configured context length is 262,144 tokens. License: apache-2.0. Quant sizes are the GGUF file bytes published on Hugging Face when a file exists.

Parameters480B / 35.0B active
Context256K
Attentiongqa
Licenseapache-2.0

62 layers · 8 KV heads · head dim 128 · GGUF repo · config

Quant file sizes

QuantFile sizeSource
BF16894.5 GBHugging Face file bytes
UD-Q8_K_XL510.6 GBHugging Face file bytes
Q8_0475.3 GBHugging Face file bytes
UD-Q6_K_XL380.0 GBHugging Face file bytes
Q6_K367.0 GBHugging Face file bytes
UD-Q5_K_XL316.6 GBHugging Face file bytes
Q5_K_M317.1 GBHugging Face file bytes
Q5_K_S307.7 GBHugging Face file bytes
UD-Q4_K_XL256.6 GBHugging Face file bytes
Q4_K_M270.1 GBHugging Face file bytes
Q4_K_S254.0 GBHugging Face file bytes
IQ4_NL252.0 GBHugging Face file bytes
IQ4_XS242.6 GBHugging Face file bytes
Q4_1279.9 GBHugging Face file bytes
Q4_0253.0 GBHugging Face file bytes
UD-Q3_K_XL198.1 GBHugging Face file bytes
Q3_K_M213.5 GBHugging Face file bytes
Q3_K_S192.7 GBHugging Face file bytes
UD-IQ3_XXS187.7 GBHugging Face file bytes
UD-Q2_K_XL167.9 GBHugging Face file bytes
Q2_K325.5 GBHugging Face file bytes
UD-IQ1_M139.4 GBHugging Face file bytes

On common machines

DeviceVerdictQuantSpeed
GeForce RTX 4060 Ti 8GBWon't fitUD-IQ1_M—
GeForce RTX 3060 12GBWon't fitUD-IQ1_M—
GeForce RTX 4060 Ti 16GBWon't fitUD-IQ1_M—
GeForce RTX 4090Won't fitUD-IQ1_M—
GeForce RTX 5090Won't fitUD-IQ1_M—
Apple M4 ProWon't fitUD-IQ1_M—
Apple M5Won't fitUD-IQ1_M—
CPU only · DDR5-5600Won't fitUD-IQ1_M—

Smallest devices that fit it fully

Run it

Commands for the GeForce RTX 4060 Ti 16GB estimate (UD-IQ1_M).

llama.cpp
llama-server -hf unsloth/Qwen3-Coder-480B-A35B-Instruct-GGUF:UD-IQ1_M -c 2048 -ngl 0
Ollama
ollama run qwen3
LM Studio
lms get unsloth/Qwen3-Coder-480B-A35B-Instruct-GGUF
# Then load quant UD-IQ1_M