Is DeepSeek-V4-Flash or GLM-4.5V 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 91.3x smaller, because of how its attention layers cache.

Model comparison

DeepSeek-V4-Flash vs GLM-4.5V

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 91.3× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	DeepSeek-V4-Flash	GLM-4.5V
Parameters	291B	108B
Architecture	deepseek4	glm4moe
Layers	43	46
Native context	1,048,576	65,536
Mixture of experts	yes, 256 experts	yes, 128 experts
Quantizations published	12	23
Smallest quantization	76.87 GiB	31.56 GiB
Q4_K_M	—	—
Licence	mit	mit

KV cache by context

the term that decides long-context viability

Context	DeepSeek-V4-Flash	GLM-4.5V	Ratio
4,096	0.06 GiB	0.72 GiB	11.41×
8,192	0.06 GiB	1.44 GiB	22.82×
16,384	0.06 GiB	2.88 GiB	45.64×
32,768	0.06 GiB	5.75 GiB	91.29×
65,536	0.06 GiB	11.50 GiB	182.57×
131,072	0.06 GiB	23.00 GiB	365.15×