TEST #744 RESULTS
04/20/2025 - 1:45 AM
ACCELERATOR
MODEL
23.2
tokens/s
generation
1.70
sec
time to first token
1063
tokens/s
prompt
244
LocalScore
HOW YOU STACK UP
Explore All ResultsGemma 3 4b It - Q6_K
SYSTEM
CPU
Intel Xeon CPU E5-2670 v3 @ 2.30GHz (haswell)
RAM
63.9GB
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
1257
tokens/s
27.4
tokens/s
850
ms
pp4096+tg256
558
tokens/s
19.2
tokens/s
7.39
sec
pp2048+tg256
905
tokens/s
19.4
tokens/s
2.31
sec
pp2048+tg768
905
tokens/s
18.1
tokens/s
2.31
sec
pp1024+tg1024
1254
tokens/s
23.5
tokens/s
853
ms
pp1280+tg3072
1145
tokens/s
17.1
tokens/s
1.16
sec
pp384+tg1152
1665
tokens/s
27.6
tokens/s
265
ms
pp64+tg1024
1495
tokens/s
29.3
tokens/s
77
ms
pp16+tg1536
383
tokens/s
27.2
tokens/s
77
ms