How much VRAM does Gemma-4-Giftige-Blume-31B-v2 need?

Q4_K_M is exactly 18,687,062,208 bytes (17.40 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

How large is Gemma-4-Giftige-Blume-31B-v2's KV cache?

6.17 GiB at 32K context with an f16 cache, computed per layer from the model's architecture — most of its layers use sliding-window attention, so a flat formula overstates this substantially. Quantizing the cache to q8_0 roughly halves it.

Nimbz · text

Gemma-4-Giftige-Blume-31B-v2

Nimbz/Gemma-4-Giftige-Blume-31B-v2

Gemma-4-Giftige-Blume-31B-v2 at Q4_K_M is exactly 18,687,062,208 bytes (17.40 GiB / 18.69 GB) — an effective 4.781 bits per weight, not the nominal 4. Its KV cache at 32K is 6.17 GiB, not the 30.00 GiB a flat formula predicts.

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

Parameters

31.3B

Architecture

gemma4

60 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
Q4_K_S	16.54 GiB	17,763,164,352	4.545	—	Nimbz
Q4_K_M	17.40 GiB	18,687,062,208	4.781	—	Nimbz
Q5_K_S	19.85 GiB	21,311,840,448	5.453	—	Nimbz
Q5_K_M	20.35 GiB	21,845,569,728	5.589	—	Nimbz
Q6_K	23.47 GiB	25,201,483,968	6.448	—	Nimbz
Q8_0	30.39 GiB	32,635,674,816	8.350	—	Nimbz

KV cache by context

computed per layer — this model uses sliding-window attention

Context	KV cache (f16)●	Flat formula	Overstated by	Full / windowed / recurrent
4,096	1.80 GiB	3.75 GiB	2.09×	10 / 50 / 0
8,192	2.42 GiB	7.50 GiB	3.10×	10 / 50 / 0
16,384	3.67 GiB	15.00 GiB	4.09×	10 / 50 / 0
32,768	6.17 GiB	30.00 GiB	4.86×	10 / 50 / 0
65,536	11.17 GiB	60.00 GiB	5.37×	10 / 50 / 0
131,072	21.17 GiB	120.00 GiB	5.67×	10 / 50 / 0

50 of 60 layers cache only a 1,024-token window rather than the full context, on a period of . Figures assume the default configuration; --swa-full disables the saving entirely.

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 Qwen3-4B 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 16.38 GiB. The real file is 17.40 GiB, because a quantization is a mixture and some tensors are always kept at higher precision. The larger discrepancy is the cache: a flat formula gives 30.00 GiB at 32K context where the real figure is 6.17 GiB, because most of this model's layers cache a fixed window rather than the whole context.

Architecture

from config.json

Layers

Attention heads

KV heads

Head dim

256

Hidden size

5376

Vocab

262,144

Sliding window

1024

SWA period

—

MLA

Experts

—

Experts per token

—

use_sliding_window

—

Questions people ask

How much VRAM does Gemma-4-Giftige-Blume-31B-v2 need?: Q4_K_M is exactly 18,687,062,208 bytes (17.40 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
How large is Gemma-4-Giftige-Blume-31B-v2's KV cache?: 6.17 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 Gemma-4-Giftige-Blume-31B-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.