Model comparison

BugTraceAI-Apex-G4-26B-Master-f16 vs Qwen3.8-27B

At Q4_K_M, BugTraceAI-Apex-G4-26B-Master-f16 is the smaller download — 16,796,011,040 bytes against 16,810,714,336.

From the file· summed bytes, KV per layer

Side by side

BugTraceAI-Apex-G4-26B-Master-f16Qwen3.8-27B
Parameters25.2B27.8B
Architecturegemma4qwen35
Layers64
Native context262,144
Mixture of expertsnono
Quantizations published2022
Smallest quantization7.98 GiB8.39 GiB
Q4_K_M15.64 GiB15.66 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
ContextBugTraceAI-Apex-G4-26B-Master-f16Qwen3.8-27BRatio
4,0960.25 GiB
8,1920.50 GiB
16,3841.00 GiB
32,7682.00 GiB
65,5364.00 GiB
131,0728.00 GiB