How much VRAM does GigaAM-v3 need?

Q4_K_M is exactly 182,497,888 bytes (0.17 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

ai-sage · audio asr

GigaAM-v3

ai-sage/GigaAM-v3

GigaAM-v3 at Q4_K_M is exactly 182,497,888 bytes (0.17 GiB / 0.18 GB) — an effective 6.561 bits per weight, not the nominal 4.

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

Parameters

223M

Architecture

gigaam

null layers

Context

—

native (config.json)

License

mit

Shipped quantizations

● exact bytes, summed from published files

Quant	Size●	Exact bytes●	Effective bpw●	Tensors●	Publisher
Q4_K_M	0.17 GiB	182,497,888	6.561	—	handy-computer
Q4_K_M	0.17 GiB	183,246,752	6.588	560	handy-computer
Q4_K_M	0.17 GiB	183,948,704	6.613	560	handy-computer
Q5_K_M	0.19 GiB	204,911,200	7.367	—	handy-computer
Q5_K_M	0.19 GiB	205,690,784	7.395	—	handy-computer
Q5_K_M	0.19 GiB	206,392,736	7.420	560	handy-computer
Q6_K	0.21 GiB	226,439,776	8.141	552	handy-computer
Q6_K	0.21 GiB	227,252,000	8.170	560	handy-computer
Q6_K	0.21 GiB	227,953,952	8.195	560	handy-computer
Q8_0	0.25 GiB	272,151,136	9.784	552	handy-computer
Q8_0	0.25 GiB	273,022,880	9.816	—	handy-computer
Q8_0	0.25 GiB	273,724,832	9.841	560	handy-computer
F16	0.42 GiB	449,098,336	16.146	552	handy-computer
F16	0.42 GiB	451,084,832	16.217	—	handy-computer
F16	0.42 GiB	452,381,408	16.264	560	handy-computer
F32	0.82 GiB	883,603,552	31.767	—	handy-computer
F32	0.83 GiB	887,573,536	31.910	—	handy-computer
F32	0.83 GiB	890,138,592	32.002	—	handy-computer

Compare with

same modality, comparable size

vs nemotron-3.5-asr-streaming-0.6b vs parakeet-tdt-0.6b-v3 vs whisper-medium vs canary-180m-flash vs whisper-large-v3-turbo vs parakeet-tdt-0.6b-v2 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 0.12 GiB. The real file is 0.17 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 GigaAM-v3 need?: Q4_K_M is exactly 182,497,888 bytes (0.17 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
Which quantization of GigaAM-v3 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.