How much VRAM does Wan2.1-T2V-1.3B need?

Q4_K_M is exactly 982,716,640 bytes (0.92 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

Wan-AI · text

Wan2.1-T2V-1.3B

Wan-AI/Wan2.1-T2V-1.3B

Wan2.1-T2V-1.3B at Q4_K_M is exactly 982,716,640 bytes (0.92 GiB / 0.98 GB) — an effective 5.540 bits per weight, not the nominal 4.

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

Parameters

1.4B

Architecture

qwen35

null layers

Context

—

native (config.json)

License

—

Shipped quantizations

● exact bytes, summed from published files

Quant	Size●	Exact bytes●	Effective bpw●	Tensors●	Publisher
Q3_K_S	0.61 GiB	654,903,520	3.692	—	samuelchristlie
Q3_K_M	0.68 GiB	729,129,184	4.111	825	samuelchristlie
Q4_0	0.81 GiB	865,581,280	4.880	825	samuelchristlie
Q4_K_S	0.83 GiB	892,565,728	5.032	—	samuelchristlie
Q4_1	0.86 GiB	926,775,520	5.225	—	samuelchristlie
Q4_K_M	0.92 GiB	982,716,640	5.540	825	samuelchristlie
Q5_K_S	0.94 GiB	1,013,774,560	5.715	—	samuelchristlie
Q3_K_S	0.95 GiB	1,024,104,128	5.774	—	calcuis
Q5_0	0.97 GiB	1,039,579,360	5.861	—	samuelchristlie
Q5_K_M	1.01 GiB	1,087,410,400	6.131	825	samuelchristlie
Q5_1	1.03 GiB	1,100,773,600	6.206	—	samuelchristlie
Q6_K	1.12 GiB	1,198,647,520	6.758	825	samuelchristlie
Q3_K_L	1.15 GiB	1,238,514,368	6.982	—	calcuis
Q4_K_S	1.29 GiB	1,385,021,120	7.809	—	calcuis
Q2_K2 shards	1.31 GiB	1,408,997,280	7.944	—	calcuis
Q8_0	1.43 GiB	1,535,768,800	8.658	825	samuelchristlie
Q5_K_S	1.46 GiB	1,572,142,784	8.863	—	calcuis
Q3_K_M2 shards	1.78 GiB	1,915,591,584	10.800	—	calcuis
Q4_K_M	2.54 GiB	2,724,723,264	15.361	427	DeepBeepMeep
Q5_K_M2 shards	2.63 GiB	2,821,321,632	15.906	—	calcuis
F16	2.65 GiB	2,840,754,400	16.016	825	samuelchristlie
Q4_K_M	4.69 GiB	5,038,065,216	28.404	—	DeepBeepMeep
Q4_14 shards	4.94 GiB	5,299,287,520	29.876	—	calcuis
Q4_K_M3 shards	5.78 GiB	6,210,013,952	—	—	calcuis
Q5_14 shards	5.84 GiB	6,270,432,736	—	—	calcuis
Q6_K4 shards	6.35 GiB	6,816,701,920	—	—	calcuis
Q4_06 shards	6.50 GiB	6,980,190,944	—	—	calcuis
BF162 shards	6.76 GiB	7,254,226,848	—	—	calcuis
F163 shards	6.99 GiB	7,508,102,752	—	—	calcuis
Q5_06 shards	7.76 GiB	8,334,721,760	—	—	calcuis
Q8_06 shards	11.37 GiB	12,204,778,208	—	—	calcuis

Compare with

same modality, comparable size

vs Qwen3-4B vs HyperCLOVAX-SEED-Text-Instruct-1.5B vs Llama-3.2-1B-Instruct vs ced-base vs gemma-3-1b-it vs Qwen3-1.7B 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 0.74 GiB. The real file is 0.92 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

—

Sliding window

none

SWA period

—

MLA

Experts

—

Experts per token

—

use_sliding_window

—

Questions people ask

How much VRAM does Wan2.1-T2V-1.3B need?: Q4_K_M is exactly 982,716,640 bytes (0.92 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
Which quantization of Wan2.1-T2V-1.3B 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.