How much VRAM does Qwen3-Embedding-8B need?

Q4_K_M is exactly 4,676,804,928 bytes (4.36 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

Qwen · embedding

Qwen3-Embedding-8B

Qwen/Qwen3-Embedding-8B

Qwen3-Embedding-8B at Q4_K_M is exactly 4,676,804,928 bytes (4.36 GiB / 4.68 GB) — an effective 4.944 bits per weight, not the nominal 4.

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

Parameters

7.6B

Architecture

qwen3

36 layers

Context

40,960

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	2.87 GiB	3,076,612,864	3.252	—	mradermacher
Q3_K_S	3.26 GiB	3,501,286,144	3.701	—	mradermacher
Q3_K_M	3.59 GiB	3,855,835,904	4.076	—	mradermacher
Q3_K_L	3.88 GiB	4,163,068,672	4.401	—	mradermacher
IQ4_XS	3.97 GiB	4,261,765,888	4.505	—	mradermacher
IQ4_XS	4.06 GiB	4,357,163,776	4.606	—	cstr
Q4_K_S	4.15 GiB	4,451,033,856	4.705	—	mradermacher
Q4_K	4.26 GiB	4,574,219,008	4.836	—	cstr
Q4_K_M	4.36 GiB	4,676,804,928	4.944	—	zenlm
Q4_K_M	4.36 GiB	4,676,805,376	4.944	—	mradermacher
Q5_K_S	4.93 GiB	5,291,991,808	5.595	—	mradermacher
Q5_K_M	5.05 GiB	5,422,342,912	5.732	—	mradermacher
Q5_K	5.07 GiB	5,442,439,936	5.754	—	cstr
Q6_K	5.79 GiB	6,214,476,544	6.570	—	mradermacher
Q8_0	7.49 GiB	8,047,102,720	8.507	—	cstr
Q8_0	7.49 GiB	8,047,106,272	8.507	—	mradermacher
F16	14.10 GiB	15,141,156,832	16.007	—	mradermacher

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 nomic-embed-text-v2-moe vs snowflake-arctic-embed-l-v2.0 vs Qwen3.5-9B-DFlash vs Qwen3-Embedding-4B vs Nemotron-3-Embed-8B-BF16 vs nomic-embed-code 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 3.96 GiB. The real file is 4.36 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

151,665

Sliding window

none

SWA period

—

MLA

Experts

—

Experts per token

—

use_sliding_window

false

Questions people ask

How much VRAM does Qwen3-Embedding-8B need?: Q4_K_M is exactly 4,676,804,928 bytes (4.36 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-Embedding-8B 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.