Qwen · text · mixture of experts

Qwen3-Coder-480B-A35B-Instruct

Qwen/Qwen3-Coder-480B-A35B-Instruct

Qwen3-Coder-480B-A35B-Instruct at Q4_K_M is exactly 290,058,826,208 bytes (270.14 GiB / 290.06 GB) — an effective 4.833 bits per weight, not the nominal 4. Its KV cache at 32K is 7.75 GiB.

From the file· summed from 6 file(s)From the file· KV per layer
Parameters
480B
total, not active
Architecture
qwen3moe
62 layers
Context
262,144
native (config.json)
License
apache-2.0

Shipped quantizations

● exact bytes, summed from published files
QuantSize●Exact bytes●Effective bpw●Tensors●Publisher
IQ2_XS4 shards123.78 GiB132,909,872,5122.215—bartowski
IQ2_S4 shards124.42 GiB133,589,786,0162.226—bartowski
UD-IQ1_M139.43 GiB149,716,201,3122.494—unsloth
Q2_K5 shards156.43 GiB167,970,700,8322.799—bartowski
Q2_K_L5 shards157.28 GiB168,882,316,8642.814—bartowski
Q2_K4 shards162.66 GiB174,656,356,5442.910—unsloth
Q2_K_L4 shards162.87 GiB174,875,144,4162.914—unsloth
IQ3_XXS5 shards176.37 GiB189,374,553,6003.155—bartowski
IQ3_XS6 shards183.57 GiB197,104,903,8083.284—bartowski
UD-IQ3_XXS5 shards187.70 GiB201,542,070,5923.358—unsloth
Q3_K_S5 shards192.70 GiB206,907,582,8163.447—unsloth
Q3_K_S6 shards194.05 GiB208,361,203,2963.472—bartowski
IQ3_M6 shards203.44 GiB218,440,017,5363.639—bartowski
Q3_K_M6 shards203.44 GiB218,441,590,4003.639—bartowski
Q3_K_L6 shards211.15 GiB226,719,180,1283.777—lmstudio-community
Q3_K_L6 shards211.15 GiB226,719,180,4163.777—bartowski
Q3_K_M5 shards213.50 GiB229,248,837,9843.820—unsloth
IQ4_XS7 shards238.95 GiB256,571,789,0564.275—bartowski
IQ4_XS6 shards242.65 GiB260,538,233,2484.341—unsloth
IQ4_NL6 shards252.01 GiB270,594,732,4484.508—unsloth
IQ4_NL7 shards252.74 GiB271,374,872,3524.521—bartowski
Q4_06 shards252.99 GiB271,646,978,4964.526—unsloth
Q4_K_S6 shards254.02 GiB272,751,129,0244.544—unsloth
Q4_07 shards256.54 GiB275,459,600,1604.590—bartowski
Q4_K_S8 shards261.48 GiB280,766,443,3924.678—bartowski
Q4_K_M6 shards270.14 GiB290,058,826,2084.833—unsloth
Q4_K_M8 shards270.86 GiB290,838,965,8244.846—lmstudio-community
Q4_K_M8 shards270.86 GiB290,838,966,1124.846—bartowski
Q4_17 shards279.86 GiB300,493,451,8405.007—unsloth
Q4_18 shards280.02 GiB300,664,106,8805.010—bartowski
Q5_K_S7 shards307.75 GiB330,440,929,8885.506—unsloth
Q5_K_S9 shards307.88 GiB330,587,205,6325.508—bartowski
Q5_K_M7 shards317.11 GiB340,493,005,4085.673—unsloth
Q5_K_M9 shards317.25 GiB340,639,281,1205.676—bartowski
Q6_K9 shards367.01 GiB394,079,320,9286.566—unsloth
Q6_K11 shards367.10 GiB394,173,790,1766.567—lmstudio-community
Q6_K11 shards367.10 GiB394,173,790,4966.567—bartowski
Q8_014 shards475.30 GiB510,351,878,4968.503—lmstudio-community
Q8_014 shards475.30 GiB510,351,878,8168.503—bartowski
Q8_011 shards475.30 GiB510,351,879,2648.503—unsloth

KV cache by context

computed per layer
ContextKV cache (f16)●Flat formulaOverstated byFull / windowed / recurrent
4,0960.97 GiB0.97 GiB—62 / 0 / 0
8,1921.94 GiB1.94 GiB—62 / 0 / 0
16,3843.88 GiB3.88 GiB—62 / 0 / 0
32,7687.75 GiB7.75 GiB—62 / 0 / 0
65,53615.50 GiB15.50 GiB—62 / 0 / 0
131,07231.00 GiB31.00 GiB—62 / 0 / 0

Compare with

same modality, comparable size
vs Qwen3-Coder-30B-A3B-Instructvs Qwen3.6-27Bvs Qwen3.8-27Bvs DeepSeek-V4-Flashvs Qwen3-30B-A3B-Thinking-2507vs Laguna-XS-2.1All alternatives →Will it fit on my card? →

Will it run on your card?

full quant x context sweep
By VRAM:8GB12GB16GB24GB32GB48GB
Radeon RX 6500 XT 4GBGeForce RTX 3050 6GBGeForce RTX 5050 8GBGeForce RTX 3080 10GBGeForce RTX 2080 Ti 11GBGeForce RTX 5070 12GBGeForce RTX 5060 Ti 16GBApple M3 Pro 18GBGeForce RTX 3080 Ti 20GBGeForce RTX 5090 D V2 24GBGeForce RTX 5090 D 32GBApple M5 Max 36GBApple M5 Pro 48GBApple M5 Max 64GBApple M3 Ultra 96GBApple M5 Max 128GBApple M2 Ultra 192GBApple M3 Ultra 256GBApple M3 Ultra 512GB

Why other calculators give a different number

A parameters × bits ÷ 8 estimate puts Q4_K_M at roughly 251.54 GiB. The real file is 270.14 GiB, because a quantization is a mixture and some tensors are always kept at higher precision.

Architecture

from config.json
Layers
62
Attention heads
96
KV heads
8
Head dim
128
Hidden size
6144
Vocab
151,936
Sliding window
none
SWA period
—
MLA
no
Experts
160
Experts per token
8
use_sliding_window
false

Questions people ask

How much VRAM does Qwen3-Coder-480B-A35B-Instruct need?
Q4_K_M is exactly 290,058,826,208 bytes (270.14 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-480B-A35B-Instruct's KV cache?
7.75 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-480B-A35B-Instruct a mixture-of-experts model?
Yes — 160 experts, 8 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-480B-A35B-Instruct 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.