How much VRAM does granite-4.1-30b-heretic need?

Q4_K_M is exactly 17,490,241,536 bytes (16.29 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

How large is granite-4.1-30b-heretic's KV cache?

8.00 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.

stuarch · text

granite-4.1-30b-heretic

stuarch/granite-4.1-30b-heretic

granite-4.1-30b-heretic at Q4_K_M is exactly 17,490,241,536 bytes (16.29 GiB / 17.49 GB) — an effective 4.847 bits per weight, not the nominal 4. Its KV cache at 32K is 8.00 GiB.

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

Parameters

28.9B

Architecture

granite

64 layers

Context

131,072

native (config.json)

License

apache-2.0

Shipped quantizations

● exact bytes, summed from published files

Quant	Size●	Exact bytes●	Effective bpw●	Tensors●	Publisher
I1-IQ1_S	5.73 GiB	6,154,021,152	1.706	—	mradermacher
I1-IQ1_M	6.30 GiB	6,764,292,384	1.875	—	mradermacher
I1-IQ2_XXS	7.25 GiB	7,781,411,104	2.157	—	mradermacher
I1-IQ2_XS	8.04 GiB	8,628,660,512	2.391	—	mradermacher
I1-IQ2_S	8.28 GiB	8,888,707,360	2.463	—	mradermacher
I1-IQ2_M	9.04 GiB	9,702,402,336	2.689	—	mradermacher
I1-Q2_K_S	9.31 GiB	9,993,972,000	2.770	—	mradermacher
Q2_K	9.99 GiB	10,723,780,608	2.972	—	mradermacher
I1-Q2_K	9.99 GiB	10,723,780,896	2.972	—	mradermacher
I1-IQ3_XXS	10.43 GiB	11,195,574,560	3.103	—	mradermacher
I1-IQ3_XS	11.12 GiB	11,938,031,904	3.309	—	mradermacher
Q3_K_S	11.71 GiB	12,569,274,368	3.483	—	mradermacher
I1-Q3_K_S	11.71 GiB	12,569,274,656	3.483	—	mradermacher
I1-IQ3_S	11.74 GiB	12,604,926,240	3.493	—	mradermacher
I1-IQ3_M	12.00 GiB	12,890,138,912	3.572	—	mradermacher
Q3_K_M	13.00 GiB	13,956,540,416	3.868	—	mradermacher
I1-Q3_K_M	13.00 GiB	13,956,540,704	3.868	—	mradermacher
Q3_K_L	14.09 GiB	15,129,896,960	4.193	—	mradermacher
I1-Q3_K_L	14.09 GiB	15,129,897,248	4.193	—	mradermacher
I1-IQ4_XS	14.44 GiB	15,501,093,152	4.296	—	mradermacher
IQ4_XS	14.59 GiB	15,668,865,024	4.343	—	mradermacher
I1-Q4_0	15.29 GiB	16,415,451,424	4.550	—	mradermacher
Q4_K_S	15.35 GiB	16,484,657,152	4.569	—	mradermacher
I1-Q4_K_S	15.35 GiB	16,484,657,440	4.569	—	mradermacher
Q4_K_M	16.29 GiB	17,490,241,536	4.847	—	mradermacher
I1-Q4_K_M	16.29 GiB	17,490,241,824	4.847	—	mradermacher
I1-Q4_1	16.88 GiB	18,126,727,456	5.024	—	mradermacher
Q5_K_S	18.54 GiB	19,905,112,064	5.517	—	mradermacher
I1-Q5_K_S	18.54 GiB	19,905,112,352	5.517	—	mradermacher
Q5_K_M	19.09 GiB	20,493,363,200	5.680	—	mradermacher
I1-Q5_K_M	19.09 GiB	20,493,363,488	5.680	—	mradermacher
Q6_K	22.06 GiB	23,684,179,968	6.564	—	mradermacher
I1-Q6_K	22.06 GiB	23,684,180,256	6.564	—	mradermacher
Q8_0	28.57 GiB	30,674,970,624	8.501	—	mradermacher

KV cache by context

computed per layer

Context	KV cache (f16)●	Flat formula	Overstated by	Full / windowed / recurrent
4,096	1.00 GiB	1.00 GiB	—	64 / 0 / 0
8,192	2.00 GiB	2.00 GiB	—	64 / 0 / 0
16,384	4.00 GiB	4.00 GiB	—	64 / 0 / 0
32,768	8.00 GiB	8.00 GiB	—	64 / 0 / 0
65,536	16.00 GiB	16.00 GiB	—	64 / 0 / 0
131,072	32.00 GiB	32.00 GiB	—	64 / 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 15.12 GiB. The real file is 16.29 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

4096

Vocab

100,352

Sliding window

none

SWA period

—

MLA

Experts

—

Experts per token

—

use_sliding_window

—

Questions people ask

How much VRAM does granite-4.1-30b-heretic need?: Q4_K_M is exactly 17,490,241,536 bytes (16.29 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
How large is granite-4.1-30b-heretic's KV cache?: 8.00 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 granite-4.1-30b-heretic 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.