Is GLM-4.7-REAP-218B-A32B 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 182.6x smaller, because of how its attention layers cache.

Model comparison

GLM-4.7-REAP-218B-A32B 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 182.6× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	GLM-4.7-REAP-218B-A32B	DeepSeek-V4-Flash
Parameters	218B	291B
Architecture	glm4moe	deepseek4
Layers	92	43
Native context	202,752	1,048,576
Mixture of experts	yes, 96 experts	yes, 256 experts
Quantizations published	45	12
Smallest quantization	44.60 GiB	76.87 GiB
Q4_K_M	122.98 GiB	—
Licence	mit	mit

KV cache by context

the term that decides long-context viability

Context	GLM-4.7-REAP-218B-A32B	DeepSeek-V4-Flash	Ratio
4,096	1.44 GiB	0.06 GiB	22.82×
8,192	2.88 GiB	0.06 GiB	45.64×
16,384	5.75 GiB	0.06 GiB	91.29×
32,768	11.50 GiB	0.06 GiB	182.57×
65,536	23.00 GiB	0.06 GiB	365.15×
131,072	46.00 GiB	0.06 GiB	730.29×