How much VRAM does DigitalAhmed-V3-qwen2.5-0.5B need?

Q4_K_M is exactly 397,808,160 bytes (0.37 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

How large is DigitalAhmed-V3-qwen2.5-0.5B's KV cache?

0.38 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.

Elhusseny · text

DigitalAhmed-V3-qwen2.5-0.5B

Elhusseny/DigitalAhmed-V3-qwen2.5-0.5B

DigitalAhmed-V3-qwen2.5-0.5B at Q4_K_M is exactly 397,808,160 bytes (0.37 GiB / 0.40 GB) — an effective 6.442 bits per weight, not the nominal 4. Its KV cache at 32K is 0.38 GiB.

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

Parameters

494M

Architecture

qwen2

24 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
I1-IQ1_S	0.29 GiB	315,829,824	5.114	—	mradermacher
I1-IQ1_M	0.30 GiB	317,974,848	5.149	—	mradermacher
I1-IQ2_XXS	0.30 GiB	321,549,888	5.207	—	mradermacher
I1-IQ2_XS	0.30 GiB	324,409,920	5.253	—	mradermacher
I1-IQ2_S	0.30 GiB	325,737,792	5.275	—	mradermacher
I1-IQ2_M	0.31 GiB	328,597,824	5.321	—	mradermacher
I1-Q2_K_S	0.31 GiB	331,049,280	5.361	—	mradermacher
I1-IQ3_XXS	0.31 GiB	333,705,024	5.404	—	mradermacher
Q3_K_S	0.32 GiB	338,263,584	5.478	—	mradermacher
I1-Q3_K_S	0.32 GiB	338,263,872	5.478	—	mradermacher
Q2_K	0.32 GiB	338,607,648	5.483	—	mradermacher
I1-IQ3_S	0.32 GiB	338,607,936	5.483	—	mradermacher
I1-Q2_K	0.32 GiB	338,607,936	5.483	—	mradermacher
I1-IQ3_XS	0.32 GiB	338,607,936	5.483	—	mradermacher
I1-IQ3_M	0.32 GiB	342,752,832	5.550	—	mradermacher
I1-IQ4_XS	0.33 GiB	349,402,944	5.658	—	mradermacher
IQ4_XS	0.33 GiB	351,445,536	5.691	—	mradermacher
I1-IQ4_NL	0.33 GiB	352,671,552	5.711	—	mradermacher
I1-Q4_0	0.33 GiB	352,972,608	5.716	—	mradermacher
Q3_K_M	0.33 GiB	355,466,784	5.756	—	mradermacher
I1-Q3_K_M	0.33 GiB	355,467,072	5.756	—	mradermacher
Q3_K_L	0.34 GiB	369,358,368	5.981	—	mradermacher
I1-Q3_K_L	0.34 GiB	369,358,656	5.981	—	mradermacher
I1-Q4_1	0.35 GiB	374,519,616	6.065	—	mradermacher
Q4_K_S	0.36 GiB	385,472,032	6.242	—	mradermacher
I1-Q4_K_S	0.36 GiB	385,472,320	6.242	—	mradermacher
Q4_K_M	0.37 GiB	397,808,160	6.442	—	mradermacher
I1-Q4_K_M	0.37 GiB	397,808,448	6.442	—	mradermacher
Q5_K_S	0.38 GiB	412,710,432	6.683	—	mradermacher
I1-Q5_K_S	0.38 GiB	412,710,720	6.683	—	mradermacher
Q5_K_M	0.39 GiB	420,086,304	6.803	—	mradermacher
I1-Q5_K_M	0.39 GiB	420,086,592	6.803	—	mradermacher
Q6_K	0.47 GiB	505,736,736	8.190	—	mradermacher
I1-Q6_K	0.47 GiB	505,737,024	8.190	—	mradermacher
Q8_0	0.49 GiB	531,068,448	8.600	—	mradermacher
F16	0.93 GiB	994,157,088	16.099	—	mradermacher

KV cache by context

computed per layer

Context	KV cache (f16)●	Flat formula	Overstated by	Full / windowed / recurrent
4,096	0.05 GiB	0.05 GiB	—	24 / 0 / 0
8,192	0.09 GiB	0.09 GiB	—	24 / 0 / 0
16,384	0.19 GiB	0.19 GiB	—	24 / 0 / 0
32,768	0.38 GiB	0.38 GiB	—	24 / 0 / 0
65,536	0.75 GiB	0.75 GiB	—	24 / 0 / 0
131,072	1.50 GiB	1.50 GiB	—	24 / 0 / 0

Compare with

same modality, comparable size

vs Llama-3.2-1B-Instruct vs ced-base vs gemma-3-1b-it vs Qwen3-1.7B vs embeddinggemma-300m vs Qwen3-0.6B 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 0.26 GiB. The real file is 0.37 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

896

Vocab

151,936

Sliding window

none

SWA period

—

MLA

Experts

—

Experts per token

—

use_sliding_window

false

Questions people ask

How much VRAM does DigitalAhmed-V3-qwen2.5-0.5B need?: Q4_K_M is exactly 397,808,160 bytes (0.37 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
How large is DigitalAhmed-V3-qwen2.5-0.5B's KV cache?: 0.38 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 DigitalAhmed-V3-qwen2.5-0.5B 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.