Models › DeepSeek-R1-Distill-Qwen-14B

DeepSeek-R1-Distill-Qwen-14B: VRAM needs and speed by GPU

At Q4_K_M, DeepSeek-R1-Distill-Qwen-14B needs about 10.2 GB of VRAM. No measured runs yet. Speeds below are estimates. Submit a run.

14.77B parameters · 131,072-token context · GGUF published Jan 20, 2025 · GGUF files on Hugging Face

Memory needed per quantization

QuantFile sizeNeeds (estimate)Smallest card it fits
IQ2_XS4.7 GB5.9 GB8 GB cards
IQ2_S5.0 GB6.2 GB8 GB cards
IQ2_M5.4 GB6.5 GB8 GB cards
Q2_K5.8 GB6.9 GB8 GB cards
IQ3_XS6.4 GB7.6 GB12 GB cards
Q2_K_L6.5 GB7.7 GB12 GB cards
Q3_K_S6.7 GB7.8 GB12 GB cards
IQ3_M6.9 GB8.1 GB12 GB cards
Q3_K_M7.3 GB8.5 GB12 GB cards
Q3_K_L7.9 GB9.1 GB12 GB cards
IQ4_XS8.1 GB9.3 GB12 GB cards
Q4_08.5 GB9.7 GB12 GB cards
IQ4_NL8.5 GB9.7 GB12 GB cards
Q4_K_S8.6 GB9.7 GB12 GB cards
Q3_K_XL8.6 GB9.8 GB12 GB cards
Q4_K_M9.0 GB10.2 GB12 GB cards
Q4_19.4 GB10.6 GB12 GB cards
Q4_K_L9.6 GB10.7 GB12 GB cards
Q5_K_S10.3 GB11.4 GB12 GB cards
Q5_K_M10.5 GB11.7 GB16 GB cards
Q5_K_L11.0 GB12.2 GB16 GB cards
Q6_K12.1 GB13.3 GB16 GB cards
Q6_K_L12.5 GB13.7 GB16 GB cards
Q8_015.7 GB16.9 GB24 GB cards
F1629.5 GB30.7 GB32 GB cards

Expected speed on popular hardware

HardwareBest quant that fitsExpected tok/s
NVIDIA RTX 3060Q5_K_S30.0 estimate
NVIDIA RTX 4060 TiQ2_K42.4 estimate
NVIDIA RTX 3090Q8_041.8 estimate
NVIDIA RTX 4090Q8_046.3 estimate
NVIDIA RTX 5090Q8_073.1 estimate
AMD Radeon RX 7900 XTXQ8_046.5 estimate
Apple M4 Pro (20-core GPU) (64 GB)Q8_014.7 estimate
Apple M3 Max (40-core GPU) (64 GB)Q8_021.5 estimate
AMD Ryzen AI Max+ 395 (Radeon 8060S)Q8_013.2 estimate

Check your own card

Too big for your card?

Run DeepSeek-R1-Distill-Qwen-14B 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.