RigBench data 2026-10-22

Home › Models › Qwen3 1.7B

Qwen3 1.7B: VRAM needs and speed by GPU

At Q4_K_M, Qwen3 1.7B needs about 1.9 GB of VRAM. We have 1 measured speed results for it.

1.72B parameters · license: apache-2.0 · GGUF files on Hugging Face

Measured speeds

ChipGeneration tok/sPrompt tok/sQuantBackendTestSource
Apple M4 Max (40-core GPU)325–MLX-4bitMLX short-chat decode (warm median)john-rocky apple-silicon-llm-bench (MIT) details

Memory needed per quantization

QuantFile sizeNeeds (estimate)Smallest card it fits
UD-IQ1_S0.5 GB1.4 GB8 GB cards
UD-IQ1_M0.6 GB1.4 GB8 GB cards
UD-IQ2_XXS0.6 GB1.4 GB8 GB cards
UD-IQ2_M0.7 GB1.5 GB8 GB cards
UD-IQ3_XXS0.8 GB1.6 GB8 GB cards
Q2_K0.8 GB1.6 GB8 GB cards
Q2_K_L0.8 GB1.6 GB8 GB cards
UD-Q2_K_XL0.8 GB1.6 GB8 GB cards
Q3_K_S0.9 GB1.7 GB8 GB cards
Q3_K_M0.9 GB1.8 GB8 GB cards
UD-Q3_K_XL1.0 GB1.8 GB8 GB cards
IQ4_XS1.0 GB1.9 GB8 GB cards
IQ4_NL1.1 GB1.9 GB8 GB cards
Q4_01.1 GB1.9 GB8 GB cards
Q4_K_S1.1 GB1.9 GB8 GB cards
Q4_K_M1.1 GB1.9 GB8 GB cards
UD-Q4_K_XL1.1 GB2.0 GB8 GB cards
Q4_11.1 GB2.0 GB8 GB cards
Q5_K_S1.2 GB2.1 GB8 GB cards
Q5_K_M1.3 GB2.1 GB8 GB cards
UD-Q5_K_XL1.3 GB2.1 GB8 GB cards
Q6_K1.4 GB2.3 GB8 GB cards
UD-Q6_K_XL1.6 GB2.5 GB8 GB cards
Q8_01.8 GB2.7 GB8 GB cards
UD-Q8_K_XL2.3 GB3.2 GB8 GB cards
BF163.4 GB4.3 GB8 GB cards

Expected speed on popular hardware

HardwareBest quant that fitsExpected tok/s
NVIDIA RTX 3060Q8_0168 estimate
NVIDIA RTX 4060 TiQ8_0133 estimate
NVIDIA RTX 3090Q8_0358 estimate
NVIDIA RTX 4090Q8_0475 estimate
NVIDIA RTX 5090Q8_0626 estimate
AMD Radeon RX 7900 XTXQ8_0398 estimate
Apple M4 Pro (20-core GPU) (64 GB)Q8_0126 estimate
Apple M3 Max (40-core GPU) (64 GB)Q8_0184 estimate
AMD Ryzen AI Max+ 395 (Radeon 8060S)Q8_0113 estimate

Check your own card

Too big for your card?

Run Qwen3 1.7B 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.