How much VRAM does whisper-small need?

Q4_K_M is exactly 171,630,656 bytes (0.16 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

openai · audio asr

whisper-small

openai/whisper-small

whisper-small at Q4_K_M is exactly 171,630,656 bytes (0.16 GiB / 0.17 GB) — an effective 5.680 bits per weight, not the nominal 4.

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

Parameters

242M

Architecture

whisper

12 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	0.08 GiB	89,673,692	2.968	—	oxide-lab
Q3_K	0.11 GiB	113,846,906	3.768	—	oxide-lab
Q4_0	0.13 GiB	138,891,040	4.596	—	oxide-lab
Q4_0	0.14 GiB	145,458,032	4.814	—	oxide-lab
Q4_K	0.14 GiB	145,458,032	4.814	—	oxide-lab
Q4_1	0.14 GiB	153,873,216	5.092	—	oxide-lab
Q4_1	0.15 GiB	160,333,856	5.306	—	oxide-lab
Q4_K_M	0.16 GiB	171,630,656	5.680	481	handy-computer
Q5_0	0.16 GiB	175,209,680	5.798	—	oxide-lab
Q5_K	0.16 GiB	175,209,680	5.798	—	oxide-lab
Q5_1	0.18 GiB	190,085,504	6.291	—	oxide-lab
Q5_K_M	0.18 GiB	193,749,056	6.412	481	handy-computer
Q6_K	0.19 GiB	206,820,806	6.845	—	oxide-lab
Q6_K	0.20 GiB	212,107,328	7.019	481	handy-computer
Q8_0	0.24 GiB	258,781,600	8.564	—	oxide-lab
Q8_0	0.25 GiB	264,464,624	8.752	—	oxide-lab
Q8_0	0.25 GiB	269,751,136	8.927	481	handy-computer
F16	0.46 GiB	492,888,480	16.312	481	handy-computer
F32	0.90 GiB	968,914,848	32.065	—	handy-computer

Compare with

same modality, comparable size

vs canary-180m-flash vs GigaAM-v3 vs SenseVoiceSmall vs whisper-base vs whisper-small.en vs moonshine-streaming-small 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 0.13 GiB. The real file is 0.16 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

51,865

Sliding window

none

SWA period

—

MLA

Experts

—

Experts per token

—

use_sliding_window

—

Questions people ask

How much VRAM does whisper-small need?: Q4_K_M is exactly 171,630,656 bytes (0.16 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
Which quantization of whisper-small 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.