Models › Qwen3.5-4B

Qwen3.5-4B: VRAM needs and speed by GPU

At Q4_K_M, Qwen3.5-4B needs about 3.7 GB of VRAM. No measured runs yet. Speeds below are estimates. Submit a run.

4.21B parameters · license: apache-2.0 · 262,144-token context · GGUF published Mar 2, 2026 · GGUF files on Hugging Face

Memory needed per quantization

QuantFile sizeNeeds (estimate)Smallest card it fits
UD-IQ2_XXS1.5 GB2.4 GB8 GB cards
UD-IQ2_M1.8 GB2.7 GB8 GB cards
UD-Q2_K_XL1.9 GB2.9 GB8 GB cards
UD-IQ3_XXS1.9 GB2.9 GB8 GB cards
Q3_K_S2.1 GB3.0 GB8 GB cards
Q3_K_M2.3 GB3.2 GB8 GB cards
UD-Q3_K_XL2.4 GB3.3 GB8 GB cards
IQ4_XS2.5 GB3.4 GB8 GB cards
IQ4_NL2.6 GB3.5 GB8 GB cards
Q4_02.6 GB3.5 GB8 GB cards
Q4_K_S2.6 GB3.5 GB8 GB cards
Q4_K_M2.7 GB3.7 GB8 GB cards
Q4_12.8 GB3.7 GB8 GB cards
UD-Q4_K_XL2.9 GB3.8 GB8 GB cards
Q5_K_S3.0 GB3.9 GB8 GB cards
Q5_K_M3.1 GB4.1 GB8 GB cards
UD-Q5_K_XL3.3 GB4.2 GB8 GB cards
Q6_K3.5 GB4.4 GB8 GB cards
UD-Q6_K_XL4.1 GB5.1 GB8 GB cards
Q8_04.5 GB5.4 GB8 GB cards
UD-Q8_K_XL6.0 GB6.9 GB8 GB cards
BF168.4 GB9.3 GB12 GB cards

Expected speed on popular hardware

HardwareBest quant that fitsExpected tok/s
NVIDIA RTX 3060Q8_068.7 estimate
NVIDIA RTX 4060 TiQ8_054.6 estimate
NVIDIA RTX 3090Q8_0146 estimate
NVIDIA RTX 4090Q8_0162 estimate
NVIDIA RTX 5090Q8_0256 estimate
AMD Radeon RX 7900 XTXQ8_0163 estimate
Apple M4 Pro (20-core GPU) (64 GB)Q8_051.6 estimate
Apple M3 Max (40-core GPU) (64 GB)Q8_075.4 estimate
AMD Ryzen AI Max+ 395 (Radeon 8060S)Q8_046.4 estimate

Check your own card

Too big for your card?

Run Qwen3.5-4B 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.