How much VRAM does moonshine-streaming-small need?

Q8_0 is exactly 198,506,848 bytes (0.18 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

How large is moonshine-streaming-small's KV cache?

0.63 GiB at 32K context with an f16 cache, computed per layer from the model's architecture. Quantizing the cache to q8_0 roughly halves it.

UsefulSensors · audio asr

moonshine-streaming-small

UsefulSensors/moonshine-streaming-small

moonshine-streaming-small at Q8_0 is exactly 198,506,848 bytes (0.18 GiB / 0.20 GB) — an effective 11.332 bits per weight, not the nominal 8. Its KV cache at 32K is 0.63 GiB.

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

Parameters

140M

Architecture

moonshine_streaming

10 layers

Context

4,096

native (config.json)

License

mit

Shipped quantizations

● exact bytes, summed from published files

Quant	Size●	Exact bytes●	Effective bpw●	Tensors●	Publisher
Q8_0	0.18 GiB	198,506,848	11.332	—	handy-computer
F16	0.26 GiB	282,092,128	16.104	—	handy-computer
F32	0.52 GiB	562,146,912	32.092	—	handy-computer

KV cache by context

computed per layer

Context	KV cache (f16)●	Flat formula	Overstated by	Full / windowed / recurrent
4,096	0.08 GiB	0.08 GiB	—	10 / 0 / 0
8,192	0.16 GiB	0.16 GiB	—	10 / 0 / 0
16,384	0.31 GiB	0.31 GiB	—	10 / 0 / 0
32,768	0.63 GiB	0.63 GiB	—	10 / 0 / 0
65,536	1.25 GiB	1.25 GiB	—	10 / 0 / 0
131,072	2.50 GiB	2.50 GiB	—	10 / 0 / 0

Compare with

same modality, comparable size

vs nemotron-3.5-asr-streaming-0.6b vs parakeet-unified-en-0.6b vs parakeet-tdt-0.6b-v3 vs whisper-medium vs canary-180m-flash vs whisper-large-v3-turbo All alternatives →Will it fit on my card? →

Why other calculators give a different number

A parameters × bits ÷ 8 estimate puts Q8_0 at roughly 0.07 GiB. The real file is 0.18 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

512

Vocab

32,768

Sliding window

none

SWA period

—

MLA

Experts

—

Experts per token

—

use_sliding_window

—

Questions people ask

How much VRAM does moonshine-streaming-small need?: Q8_0 is exactly 198,506,848 bytes (0.18 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
How large is moonshine-streaming-small's KV cache?: 0.63 GiB at 32K context with an f16 cache, computed per layer. Quantizing the cache to q8_0 roughly halves it, which is often the difference between a context length fitting and not.
Which quantization of moonshine-streaming-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.