Models › Qwen3-32B

Qwen3-32B: VRAM needs and speed by GPU

At Q4_K_M, Qwen3-32B needs about 21.4 GB of VRAM. No measured runs yet. Speeds below are estimates. Submit a run.

32.76B parameters · license: apache-2.0 · 40,960-token context · GGUF published Apr 28, 2025 · GGUF files on Hugging Face

Memory needed per quantization

QuantFile sizeNeeds (estimate)Smallest card it fits
UD-IQ1_S7.7 GB9.4 GB12 GB cards
UD-IQ1_M8.3 GB9.9 GB12 GB cards
UD-IQ2_XXS9.3 GB10.9 GB12 GB cards
UD-IQ2_M11.6 GB13.2 GB16 GB cards
Q2_K12.3 GB14.0 GB16 GB cards
Q2_K_L12.5 GB14.1 GB16 GB cards
UD-Q2_K_XL12.8 GB14.4 GB16 GB cards
UD-IQ3_XXS13.0 GB14.6 GB16 GB cards
Q3_K_S14.4 GB16.0 GB24 GB cards
Q3_K_M16.0 GB17.6 GB24 GB cards
UD-Q3_K_XL16.4 GB18.0 GB24 GB cards
IQ4_XS17.7 GB19.3 GB24 GB cards
IQ4_NL18.7 GB20.3 GB24 GB cards
Q4_018.7 GB20.3 GB24 GB cards
Q4_K_S18.8 GB20.4 GB24 GB cards
Q4_K_M19.8 GB21.4 GB24 GB cards
UD-Q4_K_XL20.0 GB21.6 GB24 GB cards
Q4_120.6 GB22.3 GB24 GB cards
Q5_K_S22.6 GB24.3 GB32 GB cards
Q5_K_M23.2 GB24.8 GB32 GB cards
UD-Q5_K_XL23.2 GB24.9 GB32 GB cards
Q6_K26.9 GB28.5 GB32 GB cards
UD-Q6_K_XL29.0 GB30.6 GB32 GB cards
Q8_034.8 GB36.4 GB48 GB cards
UD-Q8_K_XL39.5 GB41.1 GB48 GB cards
BF1665.5 GB67.2 GB96 GB cards

Expected speed on popular hardware

HardwareBest quant that fitsExpected tok/s
NVIDIA RTX 3060UD-IQ2_XXS33.2 estimate
NVIDIA RTX 4060 TiDoes not fit
NVIDIA RTX 3090Q4_K_M33.2 estimate
NVIDIA RTX 4090Q4_K_M36.8 estimate
NVIDIA RTX 5090Q6_K42.7 estimate
AMD Radeon RX 7900 XTXQ4_K_M36.9 estimate
Apple M4 Pro (20-core GPU) (64 GB)Q8_06.6 estimate
Apple M3 Max (40-core GPU) (64 GB)Q8_09.7 estimate
AMD Ryzen AI Max+ 395 (Radeon 8060S)Q8_06.0 estimate

Check your own card

Too big for your card?

Run Qwen3-32B on a rented 24-80 GB GPU by the hour.

Rent a bigger GPU on RunPod

Affiliate link: if you sign up, RigBench may earn a referral credit. It never affects rankings or numbers.