How much VRAM does calme-2.3-rys-78b need?

Q4_K_M is exactly 50,699,421,824 bytes (47.22 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

How large is calme-2.3-rys-78b's KV cache?

10.75 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.

MaziyarPanahi · text

calme-2.3-rys-78b

MaziyarPanahi/calme-2.3-rys-78b

calme-2.3-rys-78b at Q4_K_M is exactly 50,699,421,824 bytes (47.22 GiB / 50.70 GB) — an effective 5.202 bits per weight, not the nominal 4. Its KV cache at 32K is 10.75 GiB.

From the file· summed from 2 file(s)From the file· KV per layer

Parameters

78.0B

Architecture

qwen2

86 layers

Context

32,768

native (config.json)

License

mit

Shipped quantizations

● exact bytes, summed from published files

Quant	Size●	Exact bytes●	Effective bpw●	Tensors●	Publisher
IQ1_M	23.68 GiB	25,421,903,648	2.608	—	bartowski
IQ2_XXS	25.43 GiB	27,302,950,688	2.801	—	bartowski
IQ2_XS	27.00 GiB	28,988,143,392	2.974	—	bartowski
IQ2_M	29.27 GiB	31,430,748,320	3.225	—	bartowski
Q2_K	29.66 GiB	31,845,779,104	3.268	—	bartowski
Q2_K_L	30.79 GiB	33,058,947,104	3.392	—	bartowski
IQ3_XXS	31.70 GiB	34,034,624,672	3.492	—	bartowski
Q3_K_S	34.24 GiB	36,766,857,248	3.773	—	bartowski
IQ3_M	35.33 GiB	37,931,956,256	3.892	—	bartowski
Q3_K_M	37.54 GiB	40,310,847,520	4.136	—	bartowski
Q3_K_L	39.44 GiB	42,346,592,288	4.345	—	bartowski
IQ4_XS	39.63 GiB	42,555,894,176	4.367	—	bartowski
Q4_K_S	43.72 GiB	46,949,189,536	4.817	—	bartowski
Q4_K_M2 shards	47.22 GiB	50,699,421,824	5.202	—	bartowski
Q4_K_L2 shards	48.08 GiB	51,621,429,472	5.297	—	bartowski
Q5_K_M2 shards	54.30 GiB	58,307,723,360	5.983	—	bartowski
Q6_K2 shards	64.27 GiB	69,014,343,680	7.082	—	bartowski
Q8_03 shards	77.16 GiB	82,851,037,536	8.501	—	bartowski

KV cache by context

computed per layer

Context	KV cache (f16)●	Flat formula	Overstated by	Full / windowed / recurrent
4,096	1.34 GiB	1.34 GiB	—	86 / 0 / 0
8,192	2.69 GiB	2.69 GiB	—	86 / 0 / 0
16,384	5.38 GiB	5.38 GiB	—	86 / 0 / 0
32,768	10.75 GiB	10.75 GiB	—	86 / 0 / 0
65,536	21.50 GiB	21.50 GiB	—	86 / 0 / 0
131,072	43.00 GiB	43.00 GiB	—	86 / 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-E4B-it vs Qwen3-30B-A3B-Thinking-2507 vs Qwen3-8B 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 40.84 GiB. The real file is 47.22 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

151,646

Sliding window

none

SWA period

—

MLA

Experts

—

Experts per token

—

use_sliding_window

false

Questions people ask

How much VRAM does calme-2.3-rys-78b need?: Q4_K_M is exactly 50,699,421,824 bytes (47.22 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
How large is calme-2.3-rys-78b's KV cache?: 10.75 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 calme-2.3-rys-78b 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.