Model comparison

gemma-4-12B-it-qat-q4_0-unquantized vs Delphi-25B-SimpleRL-Math

These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: gemma-4-12B-it-qat-q4_0-unquantized's KV cache at 32K is 13.6× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

gemma-4-12B-it-qat-q4_0-unquantizedDelphi-25B-SimpleRL-Math
Parameters12.0B25.0B
Architecturegemma4qwen3
Layers4851
Native context262,14416,384
Mixture of expertsnono
Quantizations published234
Smallest quantization6.50 GiB5.31 GiB
Q4_K_M14.12 GiB
Licenceapache-2.0

KV cache by context

the term that decides long-context viability
Contextgemma-4-12B-it-qat-q4_0-unquantizedDelphi-25B-SimpleRL-MathRatio
4,0960.72 GiB4.18 GiB5.82×
8,1920.97 GiB8.37 GiB8.64×
16,3841.47 GiB16.73 GiB11.39×
32,7682.47 GiB33.47 GiB13.56×
65,5364.47 GiB66.94 GiB14.98×
131,0728.47 GiB133.88 GiB15.81×