How much VRAM does DeepSeek-R1-Distill-Qwen-1.5B-uncensored need?

Q4_K_M is exactly 1,117,321,056 bytes (1.04 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

How large is DeepSeek-R1-Distill-Qwen-1.5B-uncensored's KV cache?

0.88 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.

thirdeyeai · text

DeepSeek-R1-Distill-Qwen-1.5B-uncensored

thirdeyeai/DeepSeek-R1-Distill-Qwen-1.5B-uncensored

DeepSeek-R1-Distill-Qwen-1.5B-uncensored at Q4_K_M is exactly 1,117,321,056 bytes (1.04 GiB / 1.12 GB) — an effective 5.030 bits per weight, not the nominal 4. Its KV cache at 32K is 0.88 GiB.

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

Parameters

1.8B

Architecture

qwen2

28 layers

Context

131,072

native (config.json)

License

mit

Shipped quantizations

● exact bytes, summed from published files

Quant	Size●	Exact bytes●	Effective bpw●	Tensors●	Publisher
I1-IQ1_S	0.48 GiB	513,103,488	2.310	—	mradermacher
I1-IQ1_M	0.50 GiB	541,037,184	2.436	—	mradermacher
I1-IQ2_XXS	0.55 GiB	587,593,344	2.645	—	mradermacher
I1-IQ2_XS	0.58 GiB	626,902,656	2.822	—	mradermacher
I1-IQ2_S	0.62 GiB	664,087,680	2.990	—	mradermacher
I1-IQ2_M	0.65 GiB	701,332,608	3.157	—	mradermacher
I1-Q2_K_S	0.67 GiB	716,711,040	3.227	—	mradermacher
Q2_K	0.70 GiB	752,880,480	3.389	—	mradermacher
I1-Q2_K	0.70 GiB	752,880,768	3.389	—	mradermacher
I1-IQ3_XXS	0.72 GiB	769,070,208	3.462	—	mradermacher
I1-IQ3_XS	0.77 GiB	831,977,088	3.745	—	mradermacher
Q3_K_S	0.80 GiB	861,222,240	3.877	—	mradermacher
I1-Q3_K_S	0.80 GiB	861,222,528	3.877	—	mradermacher
I1-IQ3_S	0.80 GiB	862,684,800	3.884	—	mradermacher
I1-IQ3_M	0.82 GiB	876,941,952	3.948	—	mradermacher
Q3_K_M	0.86 GiB	924,456,288	4.162	—	mradermacher
I1-Q3_K_M	0.86 GiB	924,456,576	4.162	—	mradermacher
Q3_K_L	0.91 GiB	980,440,416	4.414	—	mradermacher
I1-Q3_K_L	0.91 GiB	980,440,704	4.414	—	mradermacher
I1-IQ4_XS	0.95 GiB	1,019,711,616	4.590	—	mradermacher
IQ4_XS	0.96 GiB	1,026,162,528	4.620	—	mradermacher
I1-IQ4_NL	0.99 GiB	1,067,604,096	4.806	—	mradermacher
I1-Q4_0	1.00 GiB	1,068,808,320	4.811	—	mradermacher
Q4_K_S	1.00 GiB	1,071,585,120	4.824	—	mradermacher
I1-Q4_K_S	1.00 GiB	1,071,585,408	4.824	—	mradermacher
Q4_K_M	1.04 GiB	1,117,321,056	5.030	—	mradermacher
I1-Q4_K_M	1.04 GiB	1,117,321,344	5.030	—	mradermacher
I1-Q4_1	1.08 GiB	1,162,700,928	5.234	—	mradermacher
Q5_K_S	1.17 GiB	1,259,173,728	5.668	—	mradermacher
I1-Q5_K_S	1.17 GiB	1,259,174,016	5.668	—	mradermacher
Q5_K_M	1.20 GiB	1,285,494,624	5.787	—	mradermacher
I1-Q5_K_M	1.20 GiB	1,285,494,912	5.787	—	mradermacher
Q6_K	1.36 GiB	1,464,179,040	6.591	—	mradermacher
I1-Q6_K	1.36 GiB	1,464,179,328	6.591	—	mradermacher
Q8_0	1.76 GiB	1,894,532,448	8.529	—	mradermacher
F16	3.32 GiB	3,560,416,608	16.028	—	mradermacher

KV cache by context

computed per layer

Context	KV cache (f16)●	Flat formula	Overstated by	Full / windowed / recurrent
4,096	0.11 GiB	0.11 GiB	—	28 / 0 / 0
8,192	0.22 GiB	0.22 GiB	—	28 / 0 / 0
16,384	0.44 GiB	0.44 GiB	—	28 / 0 / 0
32,768	0.88 GiB	0.88 GiB	—	28 / 0 / 0
65,536	1.75 GiB	1.75 GiB	—	28 / 0 / 0
131,072	3.50 GiB	3.50 GiB	—	28 / 0 / 0

Compare with

same modality, comparable size

vs Qwen3-4B vs HyperCLOVAX-SEED-Text-Instruct-1.5B vs Llama-3.2-1B-Instruct vs ced-base vs gemma-3-1b-it vs Qwen3-1.7B 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 0.93 GiB. The real file is 1.04 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

1536

Vocab

151,936

Sliding window

none

SWA period

—

MLA

Experts

—

Experts per token

—

use_sliding_window

false

Questions people ask

How much VRAM does DeepSeek-R1-Distill-Qwen-1.5B-uncensored need?: Q4_K_M is exactly 1,117,321,056 bytes (1.04 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
How large is DeepSeek-R1-Distill-Qwen-1.5B-uncensored's KV cache?: 0.88 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 DeepSeek-R1-Distill-Qwen-1.5B-uncensored 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.