How much VRAM does GLM-Z1-32B-0414-uncensored-heretic-v2 need?

Q4_K_M is exactly 19,686,117,248 bytes (18.33 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

How large is GLM-Z1-32B-0414-uncensored-heretic-v2's KV cache?

1.91 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.

llmfan46 · text

GLM-Z1-32B-0414-uncensored-heretic-v2

llmfan46/GLM-Z1-32B-0414-uncensored-heretic-v2

GLM-Z1-32B-0414-uncensored-heretic-v2 at Q4_K_M is exactly 19,686,117,248 bytes (18.33 GiB / 19.69 GB) — an effective 4.836 bits per weight, not the nominal 4. Its KV cache at 32K is 1.91 GiB.

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

Parameters

32.6B

Architecture

glm4

61 layers

Context

32,768

native (config.json)

License

mit

Shipped quantizations

● exact bytes, summed from published files

Quant	Size●	Exact bytes●	Effective bpw●	Tensors●	Publisher
Q3_K_M	14.80 GiB	15,890,562,944	3.904	—	llmfan46
Q3_K_L	16.03 GiB	17,216,290,688	4.229	—	llmfan46
Q4_K_S	17.42 GiB	18,708,852,608	4.596	—	llmfan46
Q4_K_M	18.33 GiB	19,686,117,248	4.836	—	llmfan46
Q5_K_S	20.98 GiB	22,527,176,576	5.534	—	llmfan46
Q5_K_M	21.52 GiB	23,103,938,432	5.676	—	llmfan46
Q6_K	24.89 GiB	26,728,492,928	6.566	—	llmfan46
Q8_0	32.24 GiB	34,615,226,240	8.503	—	llmfan46
BF16	60.67 GiB	65,144,516,480	16.003	—	llmfan46

KV cache by context

computed per layer

Context	KV cache (f16)●	Flat formula	Overstated by	Full / windowed / recurrent
4,096	0.24 GiB	0.24 GiB	—	61 / 0 / 0
8,192	0.48 GiB	0.48 GiB	—	61 / 0 / 0
16,384	0.95 GiB	0.95 GiB	—	61 / 0 / 0
32,768	1.91 GiB	1.91 GiB	—	61 / 0 / 0
65,536	3.81 GiB	3.81 GiB	—	61 / 0 / 0
131,072	7.63 GiB	7.63 GiB	—	61 / 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 17.06 GiB. The real file is 18.33 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

6144

Vocab

151,552

Sliding window

none

SWA period

—

MLA

Experts

—

Experts per token

—

use_sliding_window

—

Questions people ask

How much VRAM does GLM-Z1-32B-0414-uncensored-heretic-v2 need?: Q4_K_M is exactly 19,686,117,248 bytes (18.33 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
How large is GLM-Z1-32B-0414-uncensored-heretic-v2's KV cache?: 1.91 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 GLM-Z1-32B-0414-uncensored-heretic-v2 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.