Models › Qwen3-Coder-Next

Qwen3-Coder-Next: VRAM needs and speed by GPU

At Q4_K_M, Qwen3-Coder-Next needs about 51.3 GB of VRAM. No measured runs yet. Speeds below are estimates. Submit a run.

79.67B parameters · license: apache-2.0 · 262,144-token context · GGUF published Feb 3, 2026 · GGUF files on Hugging Face

Memory needed per quantization

QuantFile sizeNeeds (estimate)Smallest card it fits
UD-TQ1_018.9 GB21.7 GB24 GB cards
UD-IQ1_S21.5 GB24.3 GB32 GB cards
UD-IQ1_M21.7 GB24.5 GB32 GB cards
UD-IQ2_XXS23.3 GB26.1 GB32 GB cards
UD-IQ2_M25.0 GB27.8 GB32 GB cards
UD-Q2_K_XL26.8 GB29.6 GB32 GB cards
UD-IQ3_XXS28.5 GB31.3 GB32 GB cards
Q2_K29.2 GB32.0 GB48 GB cards
Q2_K_L29.3 GB32.1 GB48 GB cards
UD-IQ3_S29.7 GB32.5 GB48 GB cards
UD-Q3_K_S33.3 GB36.1 GB48 GB cards
Q3_K_S34.6 GB37.4 GB48 GB cards
UD-Q3_K_M35.9 GB38.7 GB48 GB cards
UD-Q3_K_XL36.3 GB39.1 GB48 GB cards
Q3_K_M38.3 GB41.1 GB48 GB cards
UD-IQ4_XS38.4 GB41.2 GB48 GB cards
UD-IQ4_NL39.2 GB42.0 GB48 GB cards
IQ4_XS42.7 GB45.5 GB48 GB cards
IQ4_NL45.1 GB47.9 GB96 GB cards
Q4_045.3 GB48.1 GB96 GB cards
Q4_K_S45.5 GB48.3 GB96 GB cards
UD-Q4_K_S46.1 GB48.9 GB96 GB cards
MXFP4_MOE48.0 GB50.8 GB96 GB cards
Q4_K_M48.5 GB51.3 GB96 GB cards
UD-Q4_K_M49.3 GB52.1 GB96 GB cards
UD-Q4_K_XL49.6 GB52.4 GB96 GB cards
Q4_150.1 GB52.8 GB96 GB cards
Q5_K_S55.1 GB57.8 GB96 GB cards
UD-Q5_K_S55.8 GB58.6 GB96 GB cards
Q5_K_M56.8 GB59.6 GB96 GB cards
UD-Q5_K_M59.2 GB62.0 GB96 GB cards
UD-Q5_K_XL59.5 GB62.3 GB96 GB cards
Q6_K65.6 GB68.4 GB96 GB cards
UD-Q6_K65.8 GB68.6 GB96 GB cards
UD-Q6_K_XL73.1 GB75.9 GB96 GB cards
Q8_084.8 GB87.6 GB96 GB cards
UD-Q8_K_XL86.3 GB89.1 GB96 GB cards
BF16159.5 GB162.2 GBmulti-GPU / large unified memory

Expected speed on popular hardware

HardwareBest quant that fitsExpected tok/s
NVIDIA RTX 3060Does not fit
NVIDIA RTX 4060 TiDoes not fit
NVIDIA RTX 3090UD-TQ1_034.6 estimate
NVIDIA RTX 4090UD-TQ1_038.4 estimate
NVIDIA RTX 5090UD-IQ3_XXS40.3 estimate
AMD Radeon RX 7900 XTXUD-TQ1_038.5 estimate
Apple M4 Pro (20-core GPU) (64 GB)IQ4_XS5.4 estimate
Apple M3 Max (40-core GPU) (64 GB)IQ4_XS7.9 estimate
AMD Ryzen AI Max+ 395 (Radeon 8060S)Q8_02.5 estimate

Check your own card

Too big for your card?

Run Qwen3-Coder-Next 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.