Is Qwen3-VL-235B-A22B-Instruct or DeepSeek-V4-Flash smaller?

Both publish quantizations; compare the table for the exact sizes.

Which uses less memory at long context?

DeepSeek-V4-Flash — its KV cache at 32K context is 93.3x smaller, because of how its attention layers cache.

Model comparison

Qwen3-VL-235B-A22B-Instruct vs DeepSeek-V4-Flash

These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: DeepSeek-V4-Flash's KV cache at 32K is 93.3× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	Qwen3-VL-235B-A22B-Instruct	DeepSeek-V4-Flash
Parameters	236B	291B
Architecture	qwen3vlmoe	deepseek4
Layers	94	43
Native context	262,144	1,048,576
Mixture of experts	yes, 128 experts	yes, 256 experts
Quantizations published	39	12
Smallest quantization	50.67 GiB	76.87 GiB
Q4_K_M	132.39 GiB	—
Licence	apache-2.0	mit

KV cache by context

the term that decides long-context viability

Context	Qwen3-VL-235B-A22B-Instruct	DeepSeek-V4-Flash	Ratio
4,096	0.73 GiB	0.06 GiB	11.66×
8,192	1.47 GiB	0.06 GiB	23.32×
16,384	2.94 GiB	0.06 GiB	46.64×
32,768	5.88 GiB	0.06 GiB	93.27×
65,536	11.75 GiB	0.06 GiB	186.54×
131,072	23.50 GiB	0.06 GiB	373.09×