How much VRAM does Pluto-Genesis-0.6B need?

Q4_K_M is exactly 396,702,944 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 Pluto-Genesis-0.6B's KV cache?

3.50 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.

Siddh07ETH · text

Pluto-Genesis-0.6B

Siddh07ETH/Pluto-Genesis-0.6B

Pluto-Genesis-0.6B at Q4_K_M is exactly 396,702,944 bytes (0.37 GiB / 0.40 GB) — an effective 5.324 bits per weight, not the nominal 4. Its KV cache at 32K is 3.50 GiB.

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

Parameters

596M

Architecture

qwen3

28 layers

Context

40,960

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	0.19 GiB	208,013,792	2.792	—	mradermacher
I1-IQ1_M	0.20 GiB	216,050,144	2.900	—	mradermacher
I1-IQ2_XXS	0.21 GiB	229,444,064	3.079	—	mradermacher
I1-IQ2_XS	0.23 GiB	241,994,208	3.248	—	mradermacher
I1-IQ2_S	0.24 GiB	254,192,096	3.412	—	mradermacher
I1-IQ2_M	0.25 GiB	264,907,232	3.555	—	mradermacher
I1-IQ3_XXS	0.26 GiB	279,013,856	3.745	—	mradermacher
I1-Q2_K_S	0.26 GiB	280,557,024	3.766	—	mradermacher
Q2_K	0.28 GiB	296,236,256	3.976	—	mradermacher
I1-Q2_K	0.28 GiB	296,236,512	3.976	—	mradermacher
I1-IQ3_XS	0.29 GiB	312,751,584	4.198	—	mradermacher
Q3_K_S	0.30 GiB	323,073,248	4.336	—	mradermacher
I1-IQ3_S	0.30 GiB	323,073,504	4.336	—	mradermacher
I1-Q3_K_S	0.30 GiB	323,073,504	4.336	—	mradermacher
I1-IQ3_M	0.31 GiB	336,025,056	4.510	—	mradermacher
Q3_K_M	0.32 GiB	347,124,960	4.659	—	mradermacher
I1-Q3_K_M	0.32 GiB	347,125,216	4.659	—	mradermacher
I1-IQ4_XS	0.34 GiB	367,801,824	4.936	—	mradermacher
Q3_K_L	0.34 GiB	368,489,696	4.946	—	mradermacher
I1-Q3_K_L	0.34 GiB	368,489,952	4.946	—	mradermacher
IQ4_XS	0.34 GiB	369,276,128	4.956	—	mradermacher
I1-IQ4_NL	0.36 GiB	381,564,384	5.121	—	mradermacher
I1-Q4_0	0.36 GiB	382,154,208	5.129	—	mradermacher
Q4_K_S	0.36 GiB	383,268,064	5.144	—	mradermacher
I1-Q4_K_S	0.36 GiB	383,268,320	5.144	—	mradermacher
Q4_K_M	0.37 GiB	396,702,944	5.324	—	mradermacher
I1-Q4_K_M	0.37 GiB	396,703,200	5.324	—	mradermacher
I1-Q4_1	0.38 GiB	409,089,504	5.491	—	mradermacher
Q5_K_S	0.41 GiB	436,614,368	5.860	—	mradermacher
I1-Q5_K_S	0.41 GiB	436,614,624	5.860	—	mradermacher
Q5_K_M	0.41 GiB	444,413,152	5.965	—	mradermacher
I1-Q5_K_M	0.41 GiB	444,413,408	5.965	—	mradermacher
Q6_K	0.46 GiB	495,105,248	6.645	—	mradermacher
I1-Q6_K	0.46 GiB	495,105,504	6.645	—	mradermacher
Q8_0	0.60 GiB	639,445,216	8.582	—	mradermacher
F16	1.12 GiB	1,198,180,576	16.082	—	mradermacher

KV cache by context

computed per layer

Context	KV cache (f16)●	Flat formula	Overstated by	Full / windowed / recurrent
4,096	0.44 GiB	0.44 GiB	—	28 / 0 / 0
8,192	0.88 GiB	0.88 GiB	—	28 / 0 / 0
16,384	1.75 GiB	1.75 GiB	—	28 / 0 / 0
32,768	3.50 GiB	3.50 GiB	—	28 / 0 / 0
65,536	7.00 GiB	7.00 GiB	—	28 / 0 / 0
131,072	14.00 GiB	14.00 GiB	—	28 / 0 / 0

Compare with

same modality, comparable size

vs Llama-3.2-1B-Instruct vs ced-base vs embeddinggemma-300m vs Qwen3-0.6B vs LFM2.5-1.2B-Instruct vs MiniCPM5-1B-Claude-Opus-Fable5-Thinking 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 0.31 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

128

Hidden size

1024

Vocab

151,936

Sliding window

none

SWA period

—

MLA

Experts

—

Experts per token

—

use_sliding_window

false

Questions people ask

How much VRAM does Pluto-Genesis-0.6B need?: Q4_K_M is exactly 396,702,944 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 Pluto-Genesis-0.6B's KV cache?: 3.50 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 Pluto-Genesis-0.6B 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.