How much VRAM does Qwen3-Coder-Next need?

Q4_K_M is exactly 48,487,031,968 bytes (45.16 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

How large is Qwen3-Coder-Next's KV cache?

3.00 GiB at 32K context with an f16 cache, computed per layer from the model's architecture. Quantizing the cache to q8_0 roughly halves it.

Is Qwen3-Coder-Next a mixture-of-experts model?

Yes — 512 experts with 10 routed per token. Every expert must be resident in memory, but only the routed ones are read per token, so its memory requirement and its speed behave very differently.

Qwen · text · mixture of experts

Qwen3-Coder-Next

Qwen/Qwen3-Coder-Next

Qwen3-Coder-Next at Q4_K_M is exactly 48,487,031,968 bytes (45.16 GiB / 48.49 GB) — an effective 4.869 bits per weight, not the nominal 4. Its KV cache at 32K is 3.00 GiB.

From the file· summed from 1 file(s)From the file· KV per layer

Parameters

79.7B

total, not active

Architecture

qwen3next

48 layers

Context

262,144

native (config.json)

License

apache-2.0

Shipped quantizations

● exact bytes, summed from published files

Quant	Size●	Exact bytes●	Effective bpw●	Tensors●	Publisher
IQ1_S	15.44 GiB	16,582,639,296	1.665	—	bartowski
IQ1_M	16.11 GiB	17,302,421,184	1.737	—	bartowski
UD-TQ1_0	17.64 GiB	18,941,835,296	1.902	—	unsloth
IQ2_XXS	17.97 GiB	19,298,975,424	1.938	—	bartowski
UD-IQ1_S	20.03 GiB	21,508,749,344	2.160	—	unsloth
UD-IQ1_M	20.21 GiB	21,701,693,440	2.179	—	unsloth
IQ2_XS	20.69 GiB	22,217,424,576	2.231	—	bartowski
UD-IQ2_XXS	21.71 GiB	23,312,490,496	2.341	—	unsloth
IQ2_S	21.76 GiB	23,368,906,432	2.346	—	bartowski
UD-IQ2_M	23.25 GiB	24,962,293,760	2.506	—	unsloth
IQ2_M	24.31 GiB	26,098,611,904	2.620	—	bartowski
Q2_K	26.23 GiB	28,167,608,000	2.828	—	bartowski
Q2_K_L	26.52 GiB	28,471,480,000	2.859	—	bartowski
UD-IQ3_XXS	26.53 GiB	28,482,506,752	2.860	—	unsloth
Q2_K	27.22 GiB	29,224,450,080	2.934	—	unsloth
Q2_K_L	27.29 GiB	29,297,379,360	2.942	—	unsloth
UD-IQ3_S	27.65 GiB	29,690,687,488	2.981	—	unsloth
IQ3_XXS	29.55 GiB	31,726,712,512	3.186	—	bartowski
IQ3_XS	30.76 GiB	33,031,048,896	3.317	—	bartowski
UD-Q3_K_S	31.03 GiB	33,320,267,776	3.346	—	unsloth
Q3_K_S	32.23 GiB	34,603,089,952	3.474	—	unsloth
Q3_K_S	32.47 GiB	34,864,221,888	3.501	—	bartowski
UD-Q3_K_M	33.47 GiB	35,937,513,472	3.608	—	unsloth
IQ3_M	34.13 GiB	36,643,917,504	3.679	—	bartowski
Q3_K_M	34.14 GiB	36,655,713,984	3.681	843	bartowski
Q3_K_L	35.60 GiB	38,227,005,120	3.838	—	bartowski
Q3_K_M	35.69 GiB	38,322,487,328	3.848	843	unsloth
UD-IQ4_XS	35.79 GiB	38,429,272,064	3.859	—	unsloth
UD-IQ4_NL	36.54 GiB	39,234,725,888	3.939	—	unsloth
IQ4_XS	39.75 GiB	42,676,135,968	4.285	843	unsloth
IQ4_XS	39.91 GiB	42,848,224,960	4.302	843	bartowski
IQ4_NL	42.03 GiB	45,130,344,480	4.532	—	unsloth
IQ4_NL	42.20 GiB	45,310,830,272	4.550	—	bartowski
Q4_0	42.22 GiB	45,330,098,208	4.551	843	unsloth
Q4_K_S	42.40 GiB	45,531,949,088	4.572	—	unsloth
UD-Q4_K_S	42.92 GiB	46,079,830,016	4.627	—	unsloth
Q4_0	42.93 GiB	46,099,621,568	4.629	843	bartowski
Q4_K_S	43.71 GiB	46,930,880,192	4.712	—	bartowski
Q4_K_M	45.16 GiB	48,487,031,968	4.869	843	lmstudio-community
Q4_K_M	45.20 GiB	48,528,320,544	4.873	843	unsloth

KV cache by context

computed per layer

Context	KV cache (f16)●	Flat formula	Overstated by	Full / windowed / recurrent
4,096	0.38 GiB	0.38 GiB	—	48 / 0 / 0
8,192	0.75 GiB	0.75 GiB	—	48 / 0 / 0
16,384	1.50 GiB	1.50 GiB	—	48 / 0 / 0
32,768	3.00 GiB	3.00 GiB	—	48 / 0 / 0
65,536	6.00 GiB	6.00 GiB	—	48 / 0 / 0
131,072	12.00 GiB	12.00 GiB	—	48 / 0 / 0

Compare with

same modality, comparable size

vs Qwen3-Coder-30B-A3B-Instruct vs Qwen3.6-27B vs Qwen3.8-27B vs gemma-4-12B-it-qat-q4_0-unquantized vs gemma-4-E4B-it vs Qwen3-30B-A3B-Thinking-2507 All alternatives →Will it fit on my card? →

Why other calculators give a different number

A parameters × bits ÷ 8 estimate puts Q4_K_M at roughly 41.74 GiB. The real file is 45.16 GiB, because a quantization is a mixture and some tensors are always kept at higher precision.

Architecture

from config.json

Layers

Attention heads

KV heads

Head dim

256

Hidden size

2048

Vocab

151,936

Sliding window

none

SWA period

—

MLA

Experts

512

Experts per token

use_sliding_window

false

Questions people ask

How much VRAM does Qwen3-Coder-Next need?: Q4_K_M is exactly 48,487,031,968 bytes (45.16 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
How large is Qwen3-Coder-Next's KV cache?: 3.00 GiB at 32K context with an f16 cache, computed per layer. Quantizing the cache to q8_0 roughly halves it, which is often the difference between a context length fitting and not.
Is Qwen3-Coder-Next a mixture-of-experts model?: Yes — 512 experts, 10 routed per token. Every expert must be resident, but only the routed ones are read per token, which is why its memory requirement and its speed behave very differently.
Which quantization of Qwen3-Coder-Next should I use?: Q4_K_M is the usual default. Pick the largest quantization that fits your card at the context you actually need — the table above gives exact sizes for every one published.