How much VRAM does t5-v1_1-xxl need?

Q4_K_M is exactly 2,896,123,072 bytes (2.70 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

google · text

t5-v1_1-xxl

google/t5-v1_1-xxl

t5-v1_1-xxl at Q4_K_M is exactly 2,896,123,072 bytes (2.70 GiB / 2.90 GB) — an effective 4.865 bits per weight, not the nominal 4.

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

Parameters

4.8B

Architecture

t5encoder

null layers

Context

—

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	1.65 GiB	1,772,312,064	2.977	—	chatpig
Q3_K_S	1.96 GiB	2,099,467,456	3.527	—	chatpig
Q3_K_S	1.96 GiB	2,099,467,456	3.527	—	city96
Q3_K_S	1.96 GiB	2,099,467,776	3.527	—	chatpig
Q3_K_M	2.14 GiB	2,296,075,456	3.857	219	city96
Q3_K_M	2.14 GiB	2,296,075,456	3.857	—	chatpig
Q3_K_M	2.14 GiB	2,296,075,776	3.857	—	chatpig
Q3_K_L	2.29 GiB	2,462,799,040	4.137	—	chatpig
Q3_K_L	2.29 GiB	2,462,799,040	4.137	—	city96
Q3_K_L	2.29 GiB	2,462,799,360	4.137	—	chatpig
IQ4_XS	2.41 GiB	2,589,414,912	4.350	—	chatpig
Q4_0	2.53 GiB	2,714,457,600	4.560	—	chatpig
IQ4_NL	2.54 GiB	2,730,186,240	4.586	—	chatpig
Q4_K_S	2.55 GiB	2,738,574,528	4.600	—	chatpig
Q4_K_S	2.55 GiB	2,738,574,528	4.600	—	city96
Q4_K_S	2.55 GiB	2,738,574,848	4.600	—	chatpig
Q4_K_M	2.70 GiB	2,896,123,072	4.865	—	city96
Q4_K_M	2.70 GiB	2,896,123,072	4.865	219	chatpig
Q4_K_M	2.70 GiB	2,896,123,392	4.865	—	chatpig
Q4_1	2.80 GiB	3,003,864,576	5.046	—	chatpig
Q4_02 shards	2.82 GiB	3,022,819,552	5.078	—	chatpig
Q5_0	3.07 GiB	3,293,267,040	5.532	—	chatpig
Q5_K_S	3.07 GiB	3,293,271,232	5.532	—	chatpig
Q5_K_S	3.07 GiB	3,293,271,232	5.532	—	city96
Q5_K_S	3.07 GiB	3,293,271,552	5.532	—	chatpig
Q5_0	3.07 GiB	3,293,271,552	5.532	—	chatpig
Q5_K_M	3.15 GiB	3,386,856,640	5.689	—	city96
Q5_K_M	3.15 GiB	3,386,856,640	5.689	219	chatpig
Q5_K_M	3.15 GiB	3,386,856,960	5.689	—	chatpig
Q5_1	3.34 GiB	3,582,678,528	6.018	—	chatpig
Q6_K	3.64 GiB	3,908,261,056	6.565	—	chatpig
Q6_K	3.64 GiB	3,908,261,056	6.565	219	city96
Q6_K	3.64 GiB	3,908,261,376	6.565	—	chatpig
Q8_0	4.71 GiB	5,061,584,064	8.503	219	chatpig
Q8_0	4.71 GiB	5,061,584,064	8.503	—	city96
Q8_0	4.71 GiB	5,061,584,384	8.503	—	chatpig
Q2_K2 shards	4.72 GiB	5,066,965,120	8.512	—	chatpig
F16	8.87 GiB	9,526,060,224	16.002	219	city96
F16	8.87 GiB	9,526,060,544	16.002	—	chatpig
BF16	8.87 GiB	9,526,060,544	16.002	—	chatpig

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 2.49 GiB. The real file is 2.70 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

—

Vocab

32,128

Sliding window

none

SWA period

—

MLA

Experts

—

Experts per token

—

use_sliding_window

—

Questions people ask

How much VRAM does t5-v1_1-xxl need?: Q4_K_M is exactly 2,896,123,072 bytes (2.70 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
Which quantization of t5-v1_1-xxl 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.