How much VRAM does DeepSeek-V2-Lite-Chat-Uncensored-Unbiased-Reasoner need?

Q4_K_M is exactly 10,361,112,672 bytes (9.65 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

Is DeepSeek-V2-Lite-Chat-Uncensored-Unbiased-Reasoner a mixture-of-experts model?

Yes — 64 experts with 6 routed per token. Every expert must be resident in memory, but only the routed ones are read per token, so its memory requirement and its speed behave very differently.

nicoboss · text · mixture of experts

DeepSeek-V2-Lite-Chat-Uncensored-Unbiased-Reasoner

nicoboss/DeepSeek-V2-Lite-Chat-Uncensored-Unbiased-Reasoner

DeepSeek-V2-Lite-Chat-Uncensored-Unbiased-Reasoner at Q4_K_M is exactly 10,361,112,672 bytes (9.65 GiB / 10.36 GB) — an effective 5.281 bits per weight, not the nominal 4. Its KV cache at 32K is 0.95 GiB.

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

Parameters

15.7B

total, not active

Architecture

deepseek2

27 layers

Context

4,096

native (config.json)

License

llama3.3

Shipped quantizations

● exact bytes, summed from published files

Quant	Size●	Exact bytes●	Effective bpw●	Tensors●	Publisher
Q2_K	5.99 GiB	6,431,408,288	3.278	—	mradermacher
Q3_K_S	6.97 GiB	7,485,233,216	3.815	—	mradermacher
Q3_K_M	7.57 GiB	8,124,176,448	4.141	—	mradermacher
Q3_K_L	7.88 GiB	8,456,968,256	4.310	—	mradermacher
IQ4_XS	8.05 GiB	8,640,999,392	4.404	—	mradermacher
Q4_K_S	8.88 GiB	9,530,304,608	4.857	—	mradermacher
Q4_K_M	9.65 GiB	10,361,112,672	5.281	—	mradermacher
Q5_K_S	10.37 GiB	11,137,370,720	5.676	—	mradermacher
Q5_K_M	11.03 GiB	11,845,626,464	6.037	—	mradermacher
Q6_K	13.10 GiB	14,065,719,936	7.169	—	mradermacher
Q8_0	15.55 GiB	16,692,029,536	8.507	—	ijohn07
Q8_0	15.55 GiB	16,692,030,016	8.507	—	mradermacher

KV cache by context

computed per layer

Context	KV cache (f16)●	Flat formula	Overstated by	Full / windowed / recurrent
4,096	0.12 GiB	1.05 GiB	8.89×	27 / 0 / 0
8,192	0.24 GiB	2.11 GiB	8.89×	27 / 0 / 0
16,384	0.47 GiB	4.22 GiB	8.89×	27 / 0 / 0
32,768	0.95 GiB	8.44 GiB	8.89×	27 / 0 / 0
65,536	1.90 GiB	16.88 GiB	8.89×	27 / 0 / 0
131,072	3.80 GiB	33.75 GiB	8.89×	27 / 0 / 0

This model uses multi-head latent attention. No V cache is allocated at all, and the K cache stores a 512-wide latent plus 64 rope dimensions — so reading num_key_value_heads from config.json and multiplying, as every calculator does, overstates the cache by well over an order of magnitude.

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 8.22 GiB. The real file is 9.65 GiB, because a quantization is a mixture and some tensors are always kept at higher precision. The larger discrepancy is the cache: this model allocates no value cache at all, so any formula reading num_key_value_heads overstates it by more than an order of magnitude.

Architecture

from config.json

Layers

Attention heads

KV heads

Head dim

192

Hidden size

2048

Vocab

100,002

Sliding window

none

SWA period

—

MLA

yes

Experts

Experts per token

use_sliding_window

—

Questions people ask

How much VRAM does DeepSeek-V2-Lite-Chat-Uncensored-Unbiased-Reasoner need?: Q4_K_M is exactly 10,361,112,672 bytes (9.65 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-V2-Lite-Chat-Uncensored-Unbiased-Reasoner's KV cache?: 0.95 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.
Is DeepSeek-V2-Lite-Chat-Uncensored-Unbiased-Reasoner a mixture-of-experts model?: Yes — 64 experts, 6 routed per token. Every expert must be resident, but only the routed ones are read per token, which is why its memory requirement and its speed behave very differently.
Which quantization of DeepSeek-V2-Lite-Chat-Uncensored-Unbiased-Reasoner 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.