Models › gemma-4-31B-it

gemma-4-31B-it: VRAM needs and speed by GPU

At Q4_K_M, gemma-4-31B-it needs about 19.9 GB of VRAM. No measured runs yet. Speeds below are estimates. Submit a run.

30.7B parameters · license: apache-2.0 · 262,144-token context · GGUF published Apr 1, 2026 · GGUF files on Hugging Face

Memory needed per quantization

QuantFile sizeNeeds (estimate)Smallest card it fits
UD-IQ2_XXS8.5 GB10.1 GB12 GB cards
UD-IQ2_M10.8 GB12.3 GB16 GB cards
UD-Q2_K_XL11.8 GB13.3 GB16 GB cards
UD-IQ3_XXS11.8 GB13.4 GB16 GB cards
Q3_K_S13.2 GB14.8 GB16 GB cards
Q3_K_M14.7 GB16.3 GB24 GB cards
UD-Q3_K_XL15.4 GB17.0 GB24 GB cards
IQ4_XS16.4 GB17.9 GB24 GB cards
IQ4_NL17.3 GB18.9 GB24 GB cards
Q4_017.3 GB18.9 GB24 GB cards
Q4_K_S17.4 GB19.0 GB24 GB cards
Q4_K_M18.3 GB19.9 GB24 GB cards
UD-Q4_K_XL18.8 GB20.4 GB24 GB cards
Q4_119.1 GB20.7 GB24 GB cards
Q5_K_S21.1 GB22.7 GB24 GB cards
Q5_K_M21.7 GB23.2 GB24 GB cards
UD-Q5_K_XL21.9 GB23.5 GB24 GB cards
Q6_K25.2 GB26.8 GB32 GB cards
UD-Q6_K_XL27.5 GB29.1 GB32 GB cards
Q8_032.6 GB34.2 GB48 GB cards
UD-Q8_K_XL35.0 GB36.6 GB48 GB cards
BF1661.4 GB63.0 GB96 GB cards

Expected speed on popular hardware

HardwareBest quant that fitsExpected tok/s
NVIDIA RTX 3060UD-IQ2_XXS36.1 estimate
NVIDIA RTX 4060 TiDoes not fit
NVIDIA RTX 3090Q5_K_M30.3 estimate
NVIDIA RTX 4090Q5_K_M33.6 estimate
NVIDIA RTX 5090Q6_K45.6 estimate
AMD Radeon RX 7900 XTXQ5_K_M33.7 estimate
Apple M4 Pro (20-core GPU) (64 GB)Q8_07.1 estimate
Apple M3 Max (40-core GPU) (64 GB)Q8_010.4 estimate
AMD Ryzen AI Max+ 395 (Radeon 8060S)Q8_06.4 estimate

Check your own card

Too big for your card?

Run gemma-4-31B-it 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.