How much VRAM does ShellWhisperer-1.5B need?

Q4_K_M is exactly 986,045,728 bytes (0.92 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

How large is ShellWhisperer-1.5B's KV cache?

0.88 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.

fableforge-ai · text

ShellWhisperer-1.5B

fableforge-ai/ShellWhisperer-1.5B

ShellWhisperer-1.5B at Q4_K_M is exactly 986,045,728 bytes (0.92 GiB / 0.99 GB) — an effective 5.110 bits per weight, not the nominal 4. Its KV cache at 32K is 0.88 GiB.

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

Parameters

1.5B

Architecture

qwen2

28 layers

Context

32,768

native (config.json)

License

apache-2.0

Shipped quantizations

● exact bytes, summed from published files

Quant	Size●	Exact bytes●	Effective bpw●	Tensors●	Publisher
IQ1_S	0.41 GiB	436,527,936	2.262	—	fableforge-ai
IQ1_S	0.41 GiB	436,527,936	2.262	—	lucaspetti
IQ2_XXS	0.48 GiB	511,014,976	2.648	—	King3Djbl
IQ2_XXS	0.48 GiB	511,017,760	2.648	—	fableforge-ai
IQ2_XXS	0.48 GiB	511,017,760	2.648	—	lucaspetti
IQ2_XS	0.51 GiB	550,327,072	2.852	—	fableforge-ai
IQ2_XS	0.51 GiB	550,327,072	2.852	—	lucaspetti
IQ2_S	0.53 GiB	563,810,080	2.922	—	lucaspetti
IQ2_S	0.53 GiB	563,810,080	2.922	—	fableforge-ai
IQ3_XXS	0.62 GiB	668,789,824	3.466	—	King3Djbl
IQ3_XXS	0.62 GiB	668,792,608	3.466	—	lucaspetti
IQ3_XXS	0.62 GiB	668,792,608	3.466	—	fableforge-ai
Q2_K	0.63 GiB	676,302,112	3.505	—	King3Djbl
Q2_K	0.63 GiB	676,304,928	3.505	—	fableforge-ai
Q2_K	0.63 GiB	676,304,928	3.505	—	lucaspetti
IQ3_XS	0.68 GiB	731,699,488	3.792	—	fableforge-ai
IQ3_XS	0.68 GiB	731,699,488	3.792	—	lucaspetti
Q3_K_S	0.71 GiB	760,944,672	3.943	—	lucaspetti
Q3_K_S	0.71 GiB	760,944,672	3.943	—	fableforge-ai
IQ3_S	0.71 GiB	762,407,200	3.951	—	lucaspetti
IQ3_S	0.71 GiB	762,407,200	3.951	—	fableforge-ai
IQ3_M	0.72 GiB	776,664,352	4.025	—	fableforge-ai
IQ3_M	0.72 GiB	776,664,352	4.025	—	lucaspetti
Q3_K_M	0.77 GiB	824,175,904	4.271	—	King3Djbl
Q3_K_M	0.77 GiB	824,178,720	4.271	—	fableforge-ai
Q3_K_M	0.77 GiB	824,178,720	4.271	—	lucaspetti
Q3_K_L	0.82 GiB	880,162,848	4.561	—	fableforge-ai
Q3_K_L	0.82 GiB	880,162,848	4.561	—	lucaspetti
IQ4_XS	0.83 GiB	895,729,216	4.642	—	King3Djbl
IQ4_XS	0.84 GiB	902,182,944	4.675	—	fableforge-ai
IQ4_XS	0.84 GiB	902,182,944	4.675	—	lucaspetti
Q4_0	0.87 GiB	934,952,224	4.845	—	King3Djbl
Q4_0	0.87 GiB	934,955,040	4.845	—	lucaspetti
Q4_0	0.87 GiB	934,955,040	4.845	—	fableforge-ai
IQ4_NL	0.87 GiB	936,331,552	4.852	—	lucaspetti
IQ4_NL	0.87 GiB	936,331,552	4.852	—	fableforge-ai
Q4_K_S	0.88 GiB	940,312,608	4.873	—	fableforge-ai
Q4_K_S	0.88 GiB	940,312,608	4.873	—	lucaspetti
Q4_K_M	0.92 GiB	986,045,728	5.110	—	King3Djbl
Q4_K_M	0.92 GiB	986,048,544	5.110	—	lucaspetti

KV cache by context

computed per layer

Context	KV cache (f16)●	Flat formula	Overstated by	Full / windowed / recurrent
4,096	0.11 GiB	0.11 GiB	—	28 / 0 / 0
8,192	0.22 GiB	0.22 GiB	—	28 / 0 / 0
16,384	0.44 GiB	0.44 GiB	—	28 / 0 / 0
32,768	0.88 GiB	0.88 GiB	—	28 / 0 / 0
65,536	1.75 GiB	1.75 GiB	—	28 / 0 / 0
131,072	3.50 GiB	3.50 GiB	—	28 / 0 / 0

Compare with

same modality, comparable size

vs Qwen3-4B vs Llama-3.2-1B-Instruct vs ced-base vs gemma-3-1b-it vs Qwen3-1.7B vs embeddinggemma-300m 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 0.81 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

128

Hidden size

1536

Vocab

151,936

Sliding window

none

SWA period

—

MLA

Experts

—

Experts per token

—

use_sliding_window

false

Questions people ask

How much VRAM does ShellWhisperer-1.5B need?: Q4_K_M is exactly 986,045,728 bytes (0.92 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
How large is ShellWhisperer-1.5B's KV cache?: 0.88 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 ShellWhisperer-1.5B 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.