TEST #831 RESULTS
05/06/2025 - 12:59 AM
ACCELERATOR
MODEL
40.1
tokens/s
generation
896
ms
time to first token
1384
tokens/s
prompt
396
LocalScore
HOW YOU STACK UP
Explore All ResultsGemma 3 12b - Q4_0
SYSTEM
CPU
Intel Xeon W-3235 CPU @ 3.30GHz (cascadelake)
RAM
41GB
OS
Windows
Kernel Release
10.0
Architecture
x86_64
Version
Cosmopolitan 3.9.7 MODE=x86_64
RUNTIME
Name
llamafile
Version
0.9.2
Commit Hash
a30b324
DETAILED RESULTS
TEST NAME
PROMPT
GENERATION
TTFT
pp1024+tg16
1678
tokens/s
42.8
tokens/s
633
ms
pp4096+tg256
1486
tokens/s
34.6
tokens/s
2.79
sec
pp2048+tg256
1583
tokens/s
39.3
tokens/s
1.32
sec
pp2048+tg768
1571
tokens/s
38.3
tokens/s
1.33
sec
pp1024+tg1024
1383
tokens/s
40.9
tokens/s
766
ms
pp1280+tg3072
1594
tokens/s
37.2
tokens/s
828
ms
pp384+tg1152
1660
tokens/s
41.8
tokens/s
256
ms
pp64+tg1024
1102
tokens/s
43.1
tokens/s
81
ms
pp16+tg1536
395
tokens/s
42.5
tokens/s
63
ms