How much VRAM does Kunoichi-DPO-v2-7B need?

Q4_K_M is exactly 4,368,438,944 bytes (4.07 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

SanjiWatsuki · text

Kunoichi-DPO-v2-7B

SanjiWatsuki/Kunoichi-DPO-v2-7B

Kunoichi-DPO-v2-7B at Q4_K_M is exactly 4,368,438,944 bytes (4.07 GiB / 4.37 GB) — an effective 4.826 bits per weight, not the nominal 4.

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

Parameters

7.2B

Architecture

llama

32 layers

Context

8,192

native (config.json)

License

cc-by-nc-4.0

Shipped quantizations

● exact bytes, summed from published files

Quant	Size●	Exact bytes●	Effective bpw●	Tensors●	Publisher
IQ1_S	1.50 GiB	1,612,101,472	1.781	—	brittlewis12
IQ1_M	1.63 GiB	1,754,445,664	1.938	—	brittlewis12
IQ2_XXS	1.85 GiB	1,991,685,984	2.200	—	brittlewis12
IQ2_XS	2.05 GiB	2,198,255,456	2.428	—	brittlewis12
IQ2_S	2.15 GiB	2,310,920,032	2.553	—	brittlewis12
IQ2_M	2.33 GiB	2,500,712,288	2.763	—	brittlewis12
Q2_K_S	2.36 GiB	2,528,925,536	2.794	—	brittlewis12
Q2_K	2.53 GiB	2,719,241,888	3.004	—	brittlewis12
IQ3_XXS	2.63 GiB	2,827,343,712	3.123	—	brittlewis12
IQ3_XS	2.81 GiB	3,018,815,328	3.335	—	brittlewis12
Q3_K_S	2.95 GiB	3,164,567,200	3.496	—	brittlewis12
IQ3_S	2.96 GiB	3,182,393,184	3.516	—	brittlewis12
IQ3_M	3.06 GiB	3,284,891,488	3.629	—	brittlewis12
Q3_K_M	3.28 GiB	3,518,985,888	3.888	—	brittlewis12
Q3_K_L	3.56 GiB	3,822,024,352	4.222	—	brittlewis12
IQ4_XS	3.64 GiB	3,907,688,288	4.317	—	brittlewis12
Q4_0	3.83 GiB	4,108,916,384	4.539	—	brittlewis12
IQ4_NL	3.84 GiB	4,125,693,792	4.558	—	brittlewis12
Q4_K_S	3.86 GiB	4,140,373,664	4.574	—	brittlewis12
Q4_K_M	4.07 GiB	4,368,438,944	4.826	—	brittlewis12
Q4_1	4.24 GiB	4,553,316,000	5.030	—	brittlewis12
Q5_K_S	4.65 GiB	4,997,715,616	5.521	—	brittlewis12
Q5_0	4.65 GiB	4,997,715,616	5.521	—	brittlewis12
Q5_K_M	4.78 GiB	5,131,409,056	5.669	—	brittlewis12
Q5_1	5.07 GiB	5,442,115,232	6.012	—	brittlewis12
Q6_K	5.53 GiB	5,942,064,800	6.564	—	brittlewis12
Q8_0	7.17 GiB	7,695,857,312	8.502	—	brittlewis12

KV cache by context

unresolved

This model declares a 4,096-token sliding window, but we could not establish which layers use it. Its architecture publishes the layout as a per-layer array inside the model file rather than as a period in config.json, and we have not yet ingested that array.

A flat context × layers × heads figure would be substantially too high, so we are not showing one. This is tracked as a known gap rather than filled with a guess.

Compare with

same modality, comparable size

vs gemma-4-E4B-it vs Qwen3-4B vs Qwen3-8B vs HyperCLOVAX-SEED-Text-Instruct-1.5B vs Llama-3.2-1B-Instruct vs Qwen3-VL-8B-Instruct-abliterated-v1 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 3.79 GiB. The real file is 4.07 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

32,000

Sliding window

4096

SWA period

—

MLA

Experts

—

Experts per token

—

use_sliding_window

—

Questions people ask

How much VRAM does Kunoichi-DPO-v2-7B need?: Q4_K_M is exactly 4,368,438,944 bytes (4.07 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
Which quantization of Kunoichi-DPO-v2-7B 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.