How much VRAM does Qwen3.5-9B-DFlash need?

Q4_K_M is exactly 765,959,872 bytes (0.71 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

z-lab · embedding

Qwen3.5-9B-DFlash

z-lab/Qwen3.5-9B-DFlash

Qwen3.5-9B-DFlash at Q4_K_M is exactly 765,959,872 bytes (0.71 GiB / 0.77 GB) — an effective 4.743 bits per weight, not the nominal 4.

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

Parameters

1.3B

Architecture

dflash

6 layers

Context

262,144

native (config.json)

License

—

Shipped quantizations

● exact bytes, summed from published files

Quant	Size●	Exact bytes●	Effective bpw●	Tensors●	Publisher
Q2_K	0.45 GiB	481,861,312	2.984	—	Anbeeld
Q3_K_M	0.58 GiB	624,139,968	3.865	—	Anbeeld
Q4_K_M	0.71 GiB	765,959,872	4.743	—	Anbeeld
Q4_K_M	0.71 GiB	765,960,832	4.743	—	GauravGosain
Q5_K_M	0.85 GiB	913,809,088	5.659	—	Anbeeld
Q6_K	1.00 GiB	1,070,898,880	6.631	—	Anbeeld
Q8_0	1.29 GiB	1,383,767,744	8.569	—	Anbeeld
Q8_0	1.29 GiB	1,383,768,224	8.569	—	AtomicChat
Q8_0	1.29 GiB	1,383,768,704	8.569	—	GauravGosain
BF16	2.42 GiB	2,594,873,024	16.069	—	Anbeeld
BF16	2.42 GiB	2,594,873,984	16.069	—	GauravGosain

No KV cache

architectural, not a gap in our data

This architecture allocates no KV cache. Encoder and embedding models process their input in one pass rather than generating token by token, so there is nothing to carry forward between steps and memory does not grow with context. Its footprint is the weights plus a working buffer, and that is the whole story.

Compare with

same modality, comparable size

vs jina-embeddings-v5-text-small vs embeddinggemma-300m-qat-q8_0-unquantized vs KaLM-embedding-multilingual-mini-instruct-v2.5 vs nomic-embed-text-v1.5 vs jina-embeddings-v5-text-nano vs nomic-embed-text-v2-moe 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 0.68 GiB. The real file is 0.71 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

248,320

Sliding window

4096

SWA period

—

MLA

Experts

—

Experts per token

—

use_sliding_window

true

Questions people ask

How much VRAM does Qwen3.5-9B-DFlash need?: Q4_K_M is exactly 765,959,872 bytes (0.71 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
Which quantization of Qwen3.5-9B-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.