How much VRAM does gte-large need?

Q4_K_M is exactly 215,891,488 bytes (0.20 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

thenlper · embedding

gte-large

thenlper/gte-large

gte-large at Q4_K_M is exactly 215,891,488 bytes (0.20 GiB / 0.22 GB) — an effective 5.153 bits per weight, not the nominal 4.

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

Parameters

335M

Architecture

bert

24 layers

Context

512

native (config.json)

License

mit

Shipped quantizations

● exact bytes, summed from published files

Quant	Size●	Exact bytes●	Effective bpw●	Tensors●	Publisher
Q2_K	0.13 GiB	144,227,872	3.443	—	ChristianAzinn
Q3_K_S	0.15 GiB	159,563,296	3.809	—	ChristianAzinn
Q3_K_M	0.17 GiB	181,452,320	4.331	—	ChristianAzinn
Q3_K_L	0.18 GiB	198,491,680	4.738	—	ChristianAzinn
Q4_0	0.19 GiB	199,671,328	4.766	—	ChristianAzinn
Q4_K_S	0.19 GiB	203,341,344	4.854	—	ChristianAzinn
Q4_K_M	0.20 GiB	215,891,488	5.153	—	ChristianAzinn
Q5_K_S	0.22 GiB	237,420,064	5.667	—	ChristianAzinn
Q5_0	0.22 GiB	237,420,064	5.667	—	ChristianAzinn
Q5_K_M	0.23 GiB	245,775,904	5.867	—	ChristianAzinn
Q6_K	0.26 GiB	277,528,096	6.625	—	ChristianAzinn
Q8_0	0.33 GiB	358,235,712	8.551	—	ChristianAzinn

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 nomic-embed-text-v1.5 vs jina-embeddings-v5-text-nano vs all-MiniLM-L6-v2 vs mxbai-embed-xsmall-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 0.18 GiB. The real file is 0.20 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

Hidden size

1024

Vocab

30,522

Sliding window

none

SWA period

—

MLA

Experts

—

Experts per token

—

use_sliding_window

—

Questions people ask

How much VRAM does gte-large need?: Q4_K_M is exactly 215,891,488 bytes (0.20 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
Which quantization of gte-large 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.