Models › gemma-4-26B-A4B-it

gemma-4-26B-A4B-it: VRAM needs and speed by GPU

At UD-Q4_K_M, gemma-4-26B-A4B-it needs about 18.4 GB of VRAM. No measured runs yet. Speeds below are estimates. Submit a run.

25.23B parameters (mixture-of-experts, ~3.9B active) · 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_XXS9.9 GB11.4 GB12 GB cards
UD-IQ2_M10.0 GB11.4 GB12 GB cards
UD-Q2_K_XL10.5 GB12.0 GB16 GB cards
UD-IQ3_S11.3 GB12.7 GB16 GB cards
UD-IQ3_XXS11.4 GB12.8 GB16 GB cards
UD-Q3_K_M12.7 GB14.2 GB16 GB cards
UD-Q3_K_XL12.9 GB14.3 GB16 GB cards
UD-IQ4_XS13.6 GB15.0 GB16 GB cards
UD-IQ4_NL13.6 GB15.0 GB16 GB cards
UD-Q4_K_S16.5 GB17.9 GB24 GB cards
MXFP4_MOE16.6 GB18.0 GB24 GB cards
UD-Q4_K_M16.9 GB18.4 GB24 GB cards
UD-Q4_K_XL17.0 GB18.4 GB24 GB cards
UD-Q5_K_S18.9 GB20.3 GB24 GB cards
UD-Q5_K_M21.2 GB22.6 GB24 GB cards
UD-Q5_K_XL21.2 GB22.6 GB24 GB cards
UD-Q6_K23.2 GB24.6 GB32 GB cards
UD-Q6_K_XL23.3 GB24.7 GB32 GB cards
Q8_026.9 GB28.3 GB32 GB cards
UD-Q8_K_XL27.6 GB29.1 GB32 GB cards
BF1650.5 GB51.9 GB96 GB cards

Expected speed on popular hardware

HardwareBest quant that fitsExpected tok/s
NVIDIA RTX 3060UD-IQ2_M100 estimate
NVIDIA RTX 4060 TiDoes not fit
NVIDIA RTX 3090UD-Q5_K_M101 estimate
NVIDIA RTX 4090UD-Q5_K_M112 estimate
NVIDIA RTX 5090Q8_0139 estimate
AMD Radeon RX 7900 XTXUD-Q5_K_M112 estimate
Apple M4 Pro (20-core GPU) (64 GB)Q8_028.0 estimate
Apple M3 Max (40-core GPU) (64 GB)Q8_040.9 estimate
AMD Ryzen AI Max+ 395 (Radeon 8060S)Q8_025.2 estimate

Check your own card

Too big for your card?

Run gemma-4-26B-A4B-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.