How much VRAM does Qwen3.5-4B-DFlash need?

Q4_K_M is exactly 381,456,064 bytes (0.36 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

How large is Qwen3.5-4B-DFlash's KV cache?

0.21 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.

z-lab · text

Qwen3.5-4B-DFlash

z-lab/Qwen3.5-4B-DFlash

Qwen3.5-4B-DFlash at Q4_K_M is exactly 381,456,064 bytes (0.36 GiB / 0.38 GB) — an effective 4.810 bits per weight, not the nominal 4. Its KV cache at 32K is 0.21 GiB, not the 0.75 GiB a flat formula predicts.

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

Parameters

634M

Architecture

dflash

6 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
Q2_K	0.23 GiB	243,707,584	3.073	—	Anbeeld
Q3_K_M	0.29 GiB	313,585,344	3.954	—	Anbeeld
Q4_K_M	0.36 GiB	381,456,064	4.810	—	Anbeeld
Q5_K_M	0.42 GiB	454,201,024	5.727	—	Anbeeld
Q6_K	0.49 GiB	531,492,544	6.702	—	Anbeeld
Q8_0	0.64 GiB	685,133,504	8.639	—	Anbeeld
Q8_0	0.64 GiB	685,133,984	8.639	—	AtomicChat
BF16	1.19 GiB	1,279,872,704	16.139	—	Anbeeld

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	0.09 GiB	0.09 GiB	—	1 / 5 / 0
8,192	0.12 GiB	0.19 GiB	1.57×	1 / 5 / 0
16,384	0.15 GiB	0.38 GiB	2.49×	1 / 5 / 0
32,768	0.21 GiB	0.75 GiB	3.52×	1 / 5 / 0
65,536	0.34 GiB	1.50 GiB	4.44×	1 / 5 / 0
131,072	0.59 GiB	3.00 GiB	5.10×	1 / 5 / 0

5 of 6 layers cache only a 4,096-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 Llama-3.2-1B-Instruct vs ced-base vs gemma-3-1b-it vs Qwen3-1.7B vs embeddinggemma-300m vs Qwen3-0.6B 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 0.33 GiB. The real file is 0.36 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 0.75 GiB at 32K context where the real figure is 0.21 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

128

Hidden size

2560

Vocab

248,320

Sliding window

4096

SWA period

—

MLA

Experts

—

Experts per token

—

use_sliding_window

true

Questions people ask

How much VRAM does Qwen3.5-4B-DFlash need?: Q4_K_M is exactly 381,456,064 bytes (0.36 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
How large is Qwen3.5-4B-DFlash's KV cache?: 0.21 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 Qwen3.5-4B-DFlash 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.