z-lab · text

Qwen3.6-35B-A3B-DFlash

z-lab/Qwen3.6-35B-A3B-DFlash

Qwen3.6-35B-A3B-DFlash at Q4_K_M is exactly 235,691,744 bytes (0.22 GiB / 0.24 GB) — an effective 4.886 bits per weight, not the nominal 4. Its KV cache at 32K is 0.21 GiB, not the 0.75 GiB a flat formula predicts.

From the file· summed from 1 file(s)From the file· KV per layer
Parameters
386M
Architecture
dflash
6 layers
Context
262,144
native (config.json)
License
apache-2.0

Shipped quantizations

● exact bytes, summed from published files
QuantSize●Exact bytes●Effective bpw●Tensors●Publisher
Q2_K0.14 GiB153,411,2963.180—Anbeeld
Q3_K_M0.18 GiB195,780,3204.059—Anbeeld
IQ4_XS0.20 GiB218,030,3684.520—williamliao
Q4_K_M0.22 GiB235,691,7444.886—Anbeeld
Q4_K_M0.22 GiB235,692,2244.886—Alittlehammmer
Q4_K_M0.22 GiB235,692,3204.886—williamliao
Q5_K_M0.26 GiB280,256,2245.810—Anbeeld
Q5_K0.26 GiB280,256,7045.810—Alittlehammmer
Q5_K_M0.26 GiB280,256,8005.810—williamliao
Q6_K0.31 GiB327,605,9846.791—Anbeeld
Q6_K0.31 GiB327,606,4646.791—Alittlehammmer
Q6_K0.31 GiB327,606,5606.791—williamliao
Q8_00.39 GiB421,060,3208.729—Anbeeld
Q8_00.39 GiB421,060,8008.729—Alittlehammmer
Q8_00.39 GiB421,060,8968.729—williamliao
BF160.73 GiB782,819,04016.228—Anbeeld
BF160.73 GiB782,819,52016.228—Alittlehammmer
F160.73 GiB782,819,61616.228—williamliao

KV cache by context

computed per layer — this model uses sliding-window attention
ContextKV cache (f16)●Flat formulaOverstated byFull / windowed / recurrent
4,0960.09 GiB0.09 GiB—1 / 5 / 0
8,1920.12 GiB0.19 GiB1.57×1 / 5 / 0
16,3840.15 GiB0.38 GiB2.49×1 / 5 / 0
32,7680.21 GiB0.75 GiB3.52×1 / 5 / 0
65,5360.34 GiB1.50 GiB4.44×1 / 5 / 0
131,0720.59 GiB3.00 GiB5.10×1 / 5 / 0

5 of 6 layers cache only a 4,096-token window rather than the full context, on a period of . Figures assume the default configuration; --swa-full disables the saving entirely.

Compare with

same modality, comparable size
vs ced-basevs embeddinggemma-300mvs Qwen3-0.6Bvs Qwen2.5-0.5B-Instructvs SmolLM2-135M-Instructvs gemma-3-270m-itAll alternatives →Will it fit on my card? →

Will it run on your card?

full quant x context sweep
By VRAM:8GB12GB16GB24GB32GB48GB
Radeon RX 6500 XT 4GBGeForce RTX 3050 6GBGeForce RTX 5050 8GBGeForce RTX 3080 10GBGeForce RTX 2080 Ti 11GBGeForce RTX 5070 12GBGeForce RTX 5060 Ti 16GBApple M3 Pro 18GBGeForce RTX 3080 Ti 20GBGeForce RTX 5090 D V2 24GBGeForce RTX 5090 D 32GBApple M5 Max 36GBApple M5 Pro 48GBApple M5 Max 64GBApple M3 Ultra 96GBApple M5 Max 128GBApple M2 Ultra 192GBApple M3 Ultra 256GBApple M3 Ultra 512GB

Why other calculators give a different number

A parameters × bits ÷ 8 estimate puts Q4_K_M at roughly 0.20 GiB. The real file is 0.22 GiB, because a quantization is a mixture and some tensors are always kept at higher precision. The larger discrepancy is the cache: a flat formula gives 0.75 GiB at 32K context where the real figure is 0.21 GiB, because most of this model's layers cache a fixed window rather than the whole context.

Architecture

from config.json
Layers
6
Attention heads
32
KV heads
8
Head dim
128
Hidden size
2048
Vocab
248,320
Sliding window
4096
SWA period
—
MLA
no
Experts
—
Experts per token
—
use_sliding_window
true

Questions people ask

How much VRAM does Qwen3.6-35B-A3B-DFlash need?
Q4_K_M is exactly 235,691,744 bytes (0.22 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
How large is Qwen3.6-35B-A3B-DFlash's KV cache?
0.21 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 Qwen3.6-35B-A3B-DFlash 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.