Models › Gemma-4-31B-StyleTune

Gemma-4-31B-StyleTune: VRAM needs and speed by GPU

At Q4_K_M, Gemma-4-31B-StyleTune needs about 22.0 GB of VRAM. No measured runs yet. Speeds below are estimates. Submit a run.

32.11B parameters · license: apache-2.0 · 262,144-token context · GGUF published Jun 13, 2026 · GGUF files on Hugging Face

Memory needed per quantization

QuantFile sizeNeeds (estimate)Smallest card it fits
IQ2_XXS11.3 GB12.9 GB16 GB cards
IQ2_XS12.0 GB13.6 GB16 GB cards
IQ2_S12.7 GB14.3 GB16 GB cards
Q2_K13.1 GB14.7 GB16 GB cards
IQ2_M13.3 GB14.9 GB16 GB cards
IQ3_XXS13.6 GB15.2 GB16 GB cards
IQ3_XS14.4 GB16.0 GB24 GB cards
Q2_K_L14.5 GB16.1 GB24 GB cards
Q3_K_S14.9 GB16.5 GB24 GB cards
IQ3_M15.7 GB17.3 GB24 GB cards
Q3_K_M16.5 GB18.1 GB24 GB cards
Q3_K_L17.4 GB19.0 GB24 GB cards
IQ4_XS17.9 GB19.5 GB24 GB cards
Q3_K_XL18.6 GB20.2 GB24 GB cards
IQ4_NL18.8 GB20.4 GB24 GB cards
Q4_018.9 GB20.5 GB24 GB cards
Q4_K_S19.0 GB20.6 GB24 GB cards
Q4_K_M20.4 GB22.0 GB24 GB cards
Q4_120.6 GB22.2 GB24 GB cards
Q4_K_L21.4 GB23.0 GB24 GB cards
Q5_K_S22.5 GB24.1 GB32 GB cards
Q5_K_M23.6 GB25.2 GB32 GB cards
Q5_K_L24.4 GB26.0 GB32 GB cards
Q6_K27.9 GB29.5 GB32 GB cards
Q6_K_L28.6 GB30.2 GB32 GB cards
Q8_034.1 GB35.7 GB48 GB cards
BF1664.2 GB65.8 GB96 GB cards

Expected speed on popular hardware

HardwareBest quant that fitsExpected tok/s
NVIDIA RTX 3060Does not fit
NVIDIA RTX 4060 TiDoes not fit
NVIDIA RTX 3090Q4_K_M32.2 estimate
NVIDIA RTX 4090Q4_K_M35.6 estimate
NVIDIA RTX 5090Q6_K41.2 estimate
AMD Radeon RX 7900 XTXQ4_K_M35.8 estimate
Apple M4 Pro (20-core GPU) (64 GB)Q8_06.8 estimate
Apple M3 Max (40-core GPU) (64 GB)Q8_09.9 estimate
AMD Ryzen AI Max+ 395 (Radeon 8060S)Q8_06.1 estimate

Check your own card

Too big for your card?

Run Gemma-4-31B-StyleTune 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.