How much VRAM does Assistant_Pepe_70B need?

Q4_K_M is exactly 42,520,393,952 bytes (39.60 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

How large is Assistant_Pepe_70B's KV cache?

10.00 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.

SicariusSicariiStuff · text

Assistant_Pepe_70B

SicariusSicariiStuff/Assistant_Pepe_70B

Assistant_Pepe_70B at Q4_K_M is exactly 42,520,393,952 bytes (39.60 GiB / 42.52 GB) — an effective 4.821 bits per weight, not the nominal 4. Its KV cache at 32K is 10.00 GiB.

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

Parameters

70.6B

Architecture

llama

80 layers

Context

131,072

native (config.json)

License

llama3.1

Shipped quantizations

● exact bytes, summed from published files

Quant	Size●	Exact bytes●	Effective bpw●	Tensors●	Publisher
I1-IQ1_S	14.29 GiB	15,343,484,416	1.740	—	mradermacher
I1-IQ1_M	15.60 GiB	16,751,197,696	1.899	—	mradermacher
I1-IQ2_XXS	17.79 GiB	19,097,386,496	2.165	—	mradermacher
IQ1_S	18.07 GiB	19,400,162,400	2.200	—	bartowski
IQ1_M	19.14 GiB	20,556,217,440	2.331	—	bartowski
I1-IQ2_XS	19.69 GiB	21,142,109,696	2.397	—	mradermacher
I1-IQ2_S	20.71 GiB	22,242,344,448	2.522	—	mradermacher
IQ2_XXS	20.94 GiB	22,482,975,840	2.549	—	bartowski
I1-IQ2_M	22.46 GiB	24,119,295,488	2.735	—	mradermacher
IQ2_XS	22.53 GiB	24,192,154,720	2.743	—	bartowski
I1-Q2_K_S	22.79 GiB	24,471,944,704	2.775	—	mradermacher
IQ2_S	23.67 GiB	25,411,664,992	2.881	—	bartowski
Q2_K	24.56 GiB	26,375,108,800	2.991	—	SicariusSicariiStuff
I1-Q2_K	24.56 GiB	26,375,110,144	2.991	—	mradermacher
IQ2_M	25.10 GiB	26,953,071,712	3.056	—	bartowski
I1-IQ3_XXS	25.58 GiB	27,469,495,808	3.115	—	mradermacher
Q2_K	25.79 GiB	27,696,315,488	3.140	—	bartowski
Q2_K_L	26.75 GiB	28,722,363,488	3.257	—	bartowski
IQ3_XXS	27.03 GiB	29,021,387,872	3.291	—	bartowski
I1-IQ3_XS	27.29 GiB	29,307,731,456	3.323	—	mradermacher
IQ3_XS	28.54 GiB	30,639,422,560	3.474	—	bartowski
I1-Q3_K_S	28.79 GiB	30,912,052,736	3.505	—	mradermacher
I1-IQ3_S	28.79 GiB	30,912,052,736	3.505	—	mradermacher
I1-IQ3_M	29.74 GiB	31,937,035,776	3.621	—	mradermacher
Q3_K_S	29.77 GiB	31,960,628,320	3.624	—	bartowski
IQ3_M	30.72 GiB	32,985,611,360	3.740	—	bartowski
Q3_K_M	31.91 GiB	34,267,494,624	3.886	—	SicariusSicariiStuff
I1-Q3_K_M	31.91 GiB	34,267,495,936	3.886	—	mradermacher
Q3_K_M	32.89 GiB	35,316,071,520	4.004	—	bartowski
I1-Q3_K_L	34.59 GiB	37,140,594,176	4.211	—	mradermacher
Q3_K_L	34.94 GiB	37,518,081,120	4.254	—	bartowski
I1-IQ4_XS	35.30 GiB	37,902,663,168	4.298	—	mradermacher
IQ4_XS	36.11 GiB	38,767,737,952	4.396	—	bartowski
I1-Q4_0	37.36 GiB	40,116,534,784	4.549	—	mradermacher
I1-Q4_K_S	37.58 GiB	40,347,221,504	4.575	—	mradermacher
Q4_0	38.01 GiB	40,808,594,528	4.627	—	bartowski
IQ4_NL	38.03 GiB	40,834,808,928	4.630	—	bartowski
Q4_K_S	38.22 GiB	41,039,281,248	4.653	—	bartowski
Q4_K_M	39.60 GiB	42,520,393,952	4.821	—	SicariusSicariiStuff
I1-Q4_K_M	39.60 GiB	42,520,395,264	4.821	—	mradermacher

KV cache by context

computed per layer

Context	KV cache (f16)●	Flat formula	Overstated by	Full / windowed / recurrent
4,096	1.25 GiB	1.25 GiB	—	80 / 0 / 0
8,192	2.50 GiB	2.50 GiB	—	80 / 0 / 0
16,384	5.00 GiB	5.00 GiB	—	80 / 0 / 0
32,768	10.00 GiB	10.00 GiB	—	80 / 0 / 0
65,536	20.00 GiB	20.00 GiB	—	80 / 0 / 0
131,072	40.00 GiB	40.00 GiB	—	80 / 0 / 0

Compare with

same modality, comparable size

vs Qwen3-Coder-30B-A3B-Instruct vs Qwen3.6-27B vs Qwen3.8-27B vs gemma-4-12B-it-qat-q4_0-unquantized vs gemma-4-E4B-it vs Qwen3-30B-A3B-Thinking-2507 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 36.96 GiB. The real file is 39.60 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

8192

Vocab

128,256

Sliding window

none

SWA period

—

MLA

Experts

—

Experts per token

—

use_sliding_window

—

Questions people ask

How much VRAM does Assistant_Pepe_70B need?: Q4_K_M is exactly 42,520,393,952 bytes (39.60 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
How large is Assistant_Pepe_70B's KV cache?: 10.00 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 Assistant_Pepe_70B 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.