Model comparison

GLM-4.7-Flash-REAP-23B-A3B vs Qwen3-Coder-30B-A3B-Instruct

At Q4_K_M, GLM-4.7-Flash-REAP-23B-A3B is the smaller download — 14,113,838,816 bytes against 18,556,689,568. At long context the gap widens: GLM-4.7-Flash-REAP-23B-A3B's KV cache at 32K is 1.8× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

GLM-4.7-Flash-REAP-23B-A3BQwen3-Coder-30B-A3B-Instruct
Parameters23.0B30.5B
Architecturedeepseek2qwen3moe
Layers4748
Native context202,752262,144
Mixture of expertsyes, 48 expertsyes, 128 experts
Quantizations published2146
Smallest quantization6.09 GiB7.46 GiB
Q4_K_M13.14 GiB17.28 GiB
Licencemitapache-2.0

KV cache by context

the term that decides long-context viability
ContextGLM-4.7-Flash-REAP-23B-A3BQwen3-Coder-30B-A3B-InstructRatio
4,0960.21 GiB0.38 GiB1.82×
8,1920.41 GiB0.75 GiB1.82×
16,3840.83 GiB1.50 GiB1.82×
32,7681.65 GiB3.00 GiB1.82×
65,5363.30 GiB6.00 GiB1.82×
131,0726.61 GiB12.00 GiB1.82×
GLM-4.7-Flash-REAP-23B-A3B vs Qwen3-Coder-30B-A3B-Instruct — size, memory and hardware fit — ossmodeldb