TEST #2048 RESULTS
12/27/2025 - 8:13 AM
ACCELERATOR
MODEL
10.4
tokens/s
generation
1.90
sec
time to first token
676
tokens/s
prompt
155
LocalScore
HOW YOU STACK UP
Explore All ResultsGemma 3 27b - Q4_0
SYSTEM
CPU
RAM
119.7GB
OS
Linux
Kernel Release
6.14.0-1015-nvidia
Architecture
aarch64
Version
Cosmopolitan 3.9.7 MODE=aarch64; #15-Ubuntu SMP PREEMPT_DYNAMIC Tue Nov 25 18:02:16 UTC 2025
RUNTIME
Name
llamafile
Version
0.9.3
Commit Hash
a30b324
DETAILED RESULTS
TEST NAME
PROMPT
GENERATION
TTFT
pp1024+tg16
880
tokens/s
11.1
tokens/s
1.26
sec
pp4096+tg256
660
tokens/s
9.9
tokens/s
6.31
sec
pp2048+tg256
796
tokens/s
9.8
tokens/s
2.67
sec
pp2048+tg768
710
tokens/s
10.3
tokens/s
2.98
sec
pp1024+tg1024
871
tokens/s
10.6
tokens/s
1.27
sec
pp1280+tg3072
840
tokens/s
10.0
tokens/s
1.62
sec
pp384+tg1152
808
tokens/s
10.4
tokens/s
570
ms
pp64+tg1024
367
tokens/s
10.8
tokens/s
271
ms
pp16+tg1536
148
tokens/s
11.0
tokens/s
196
ms