How much VRAM does Qwen3.6-27B-Heretic2-Uncensored-Finetune-Thinking need?

Q4_K_M is exactly 16,861,398,400 bytes (15.70 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

DavidAU · vision language

Qwen3.6-27B-Heretic2-Uncensored-Finetune-Thinking

DavidAU/Qwen3.6-27B-Heretic2-Uncensored-Finetune-Thinking

Qwen3.6-27B-Heretic2-Uncensored-Finetune-Thinking at Q4_K_M is exactly 16,861,398,400 bytes (15.70 GiB / 16.86 GB) — an effective 4.931 bits per weight, not the nominal 4. Its KV cache at 32K is 2.00 GiB.

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

Parameters

27.4B

Architecture

qwen35

64 layers

Context

262,144

native (config.json)

License

apache-2.0

Shipped quantizations

● exact bytes, summed from published files

Quant	Size●	Exact bytes●	Effective bpw●	Tensors●	Publisher
IQ2_M	9.77 GiB	10,486,343,040	3.067	—	DavidAU
IQ2_M	9.77 GiB	10,486,343,040	3.067	—	pupsipups
IQ3_M	12.01 GiB	12,891,186,560	3.770	—	DavidAU
IQ3_M	12.01 GiB	12,891,186,560	3.770	—	pupsipups
IQ4_XS	14.34 GiB	15,397,242,240	4.503	851	pupsipups
IQ4_XS	14.34 GiB	15,397,242,240	4.503	851	DavidAU
Q4_K_S	14.81 GiB	15,900,312,960	4.650	—	pupsipups
Q4_K_S	14.81 GiB	15,900,312,960	4.650	—	DavidAU
IQ4_NL	15.01 GiB	16,115,926,400	4.713	—	DavidAU
IQ4_NL	15.01 GiB	16,115,926,400	4.713	—	pupsipups
Q4_K_M	15.70 GiB	16,861,398,400	4.931	851	DavidAU
Q4_K_M	15.70 GiB	16,861,398,400	4.931	851	pupsipups
Q5_K_S	17.69 GiB	18,990,663,040	5.553	—	pupsipups
Q5_K_S	17.69 GiB	18,990,663,040	5.553	—	DavidAU
Q5_K_M	18.20 GiB	19,542,148,480	5.715	851	DavidAU
Q5_K_M	18.20 GiB	19,542,148,480	5.715	851	pupsipups
Q6_K	20.57 GiB	22,082,528,448	6.458	—	BugTraceAI
Q6_K	20.86 GiB	22,396,159,360	6.549	851	DavidAU
Q6_K	20.86 GiB	22,396,159,360	6.549	851	pupsipups
Q8_0	27.82 GiB	29,866,341,760	8.734	851	DavidAU
Q8_0	27.82 GiB	29,866,341,760	8.734	851	pupsipups

KV cache by context

computed per layer

Context	KV cache (f16)●	Flat formula	Overstated by	Full / windowed / recurrent
4,096	0.25 GiB	1.00 GiB	4.00×	16 / 0 / 48
8,192	0.50 GiB	2.00 GiB	4.00×	16 / 0 / 48
16,384	1.00 GiB	4.00 GiB	4.00×	16 / 0 / 48
32,768	2.00 GiB	8.00 GiB	4.00×	16 / 0 / 48
65,536	4.00 GiB	16.00 GiB	4.00×	16 / 0 / 48
131,072	8.00 GiB	32.00 GiB	4.00×	16 / 0 / 48

48 of 64 layers use linear attention, which keeps a fixed-size recurrent state instead of a per-token cache. Those layers do not grow with context at all — treating them as ordinary attention, as a flat formula does, overstates this model's cache by roughly 4.0× at long context.

Compare with

same modality, comparable size

vs Qwen3.6-35B-A3B vs Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-MTP vs Qwen3.5-9B vs gemma-4-26B-A4B-it vs gemma-4-12B-it vs Qwen3.5-4B 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 14.33 GiB. The real file is 15.70 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

256

Hidden size

5120

Vocab

248,320

Sliding window

none

SWA period

—

MLA

Experts

—

Experts per token

—

use_sliding_window

—

Questions people ask

How much VRAM does Qwen3.6-27B-Heretic2-Uncensored-Finetune-Thinking need?: Q4_K_M is exactly 16,861,398,400 bytes (15.70 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
How large is Qwen3.6-27B-Heretic2-Uncensored-Finetune-Thinking's KV cache?: 2.00 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 Qwen3.6-27B-Heretic2-Uncensored-Finetune-Thinking 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.