How much VRAM does Qwen2.5-72B-Instruct need?

Q4_K_M is exactly 44,010,465,472 bytes (40.99 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

How large is Qwen2.5-72B-Instruct's KV cache?

10.00 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.

Qwen · text

Qwen2.5-72B-Instruct

Qwen/Qwen2.5-72B-Instruct

Qwen2.5-72B-Instruct at Q4_K_M is exactly 44,010,465,472 bytes (40.99 GiB / 44.01 GB) — an effective 4.843 bits per weight, not the nominal 4. Its KV cache at 32K is 10.00 GiB.

From the file· summed from 12 file(s)From the file· KV per layer

Parameters

72.7B

Architecture

qwen2

80 layers

Context

32,768

native (config.json)

License

other

Shipped quantizations

● exact bytes, summed from published files

Quant	Size●	Exact bytes●	Effective bpw●	Tensors●	Publisher
IQ1_M	22.11 GiB	23,740,212,896	2.612	—	bartowski
IQ2_XXS	23.74 GiB	25,490,024,096	2.805	—	bartowski
IQ2_XS	25.20 GiB	27,057,645,216	2.977	—	bartowski
Q2_K7 shards	25.45 GiB	27,329,259,168	3.007	—	Qwen
IQ2_M	27.32 GiB	29,338,986,144	3.228	—	bartowski
Q2_K	27.76 GiB	29,811,762,848	3.280	—	bartowski
Q2_K_L	28.90 GiB	31,028,274,848	3.414	—	bartowski
IQ3_XXS	29.66 GiB	31,845,082,784	3.504	—	bartowski
Q3_K_S	32.12 GiB	34,487,789,216	3.795	—	bartowski
Q3_K_M9 shards	33.04 GiB	35,472,140,224	3.903	—	Qwen
IQ3_M	33.07 GiB	35,503,597,216	3.906	—	bartowski
Q3_K_M	35.11 GiB	37,698,725,536	4.148	963	bartowski
Q3_K_L	36.79 GiB	39,505,225,088	4.347	—	lmstudio-community
Q3_K_L	36.79 GiB	39,505,225,376	4.347	—	bartowski
IQ4_XS	36.98 GiB	39,709,075,104	4.369	963	bartowski
Q4_011 shards	38.53 GiB	41,373,296,800	4.552	—	Qwen
Q4_0	38.54 GiB	41,383,126,688	4.553	963	bartowski
Q4_K_M12 shards	40.99 GiB	44,010,465,472	4.843	—	Qwen
Q4_K_M	44.16 GiB	47,415,715,200	5.217	963	lmstudio-community
Q4_K_M	44.16 GiB	47,415,715,488	5.217	963	bartowski
Q5_013 shards	46.88 GiB	50,337,049,024	5.539	—	Qwen
Q5_K_M14 shards	48.13 GiB	51,673,983,488	5.686	—	Qwen
Q5_K_M2 shards	50.71 GiB	54,447,466,336	5.991	—	bartowski
Q6_K16 shards	55.75 GiB	59,861,035,808	6.587	—	Qwen
Q6_K2 shards	59.93 GiB	64,347,629,120	7.080	—	lmstudio-community
Q6_K2 shards	59.93 GiB	64,347,629,408	7.080	—	bartowski
Q8_02 shards	71.96 GiB	77,262,612,064	8.501	—	lmstudio-community
Q8_02 shards	71.96 GiB	77,262,612,320	8.501	—	bartowski
Q8_021 shards	72.21 GiB	77,530,000,800	8.531	—	Qwen

KV cache by context

computed per layer

Context	KV cache (f16)●	Flat formula	Overstated by	Full / windowed / recurrent
4,096	1.25 GiB	1.25 GiB	—	80 / 0 / 0
8,192	2.50 GiB	2.50 GiB	—	80 / 0 / 0
16,384	5.00 GiB	5.00 GiB	—	80 / 0 / 0
32,768	10.00 GiB	10.00 GiB	—	80 / 0 / 0
65,536	20.00 GiB	20.00 GiB	—	80 / 0 / 0
131,072	40.00 GiB	40.00 GiB	—	80 / 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-E4B-it vs Qwen3-30B-A3B-Thinking-2507 vs Qwen3-8B 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 38.09 GiB. The real file is 40.99 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

128

Hidden size

8192

Vocab

152,064

Sliding window

131072

SWA period

—

MLA

Experts

—

Experts per token

—

use_sliding_window

false

This model declares a sliding window but sets use_sliding_window: false, so the window is not applied. Honouring the field without the flag understates KV for the whole family.

Questions people ask

How much VRAM does Qwen2.5-72B-Instruct need?: Q4_K_M is exactly 44,010,465,472 bytes (40.99 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
How large is Qwen2.5-72B-Instruct's KV cache?: 10.00 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.
Which quantization of Qwen2.5-72B-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.