TEST #3432 RESULTS
05/03/2026 - 10:12 PM
ACCELERATOR
MODEL
24.6
tokens/s
generation
1.31
sec
time to first token
1181
tokens/s
prompt
281
LocalScore
HOW YOU STACK UP
Explore All ResultsGemma 3 Finetune - Q8_0
SYSTEM
CPU
AMD Ryzen 7 5700G with Radeon Graphics (znver3)
RAM
15GB
OS
Linux
Kernel Release
6.19.13-arch1-1
Architecture
x86_64
Version
Cosmopolitan 3.9.7 MODE=x86_64; #1 SMP PREEMPT_DYNAMIC Tue, 21 Apr 2026 23:38:22 +0000
RUNTIME
Name
llamafile
Version
0.9.3
Commit Hash
a30b324
DETAILED RESULTS
TEST NAME
PROMPT
GENERATION
TTFT
pp1024+tg16
1458
tokens/s
30.5
tokens/s
735
ms
pp4096+tg256
812
tokens/s
5.0
tokens/s
5.24
sec
pp2048+tg256
1186
tokens/s
22.9
tokens/s
1.77
sec
pp2048+tg768
1110
tokens/s
21.7
tokens/s
1.89
sec
pp1024+tg1024
1438
tokens/s
26.9
tokens/s
745
ms
pp1280+tg3072
1347
tokens/s
15.8
tokens/s
986
ms
pp384+tg1152
1665
tokens/s
31.5
tokens/s
259
ms
pp64+tg1024
1142
tokens/s
34.9
tokens/s
81
ms
pp16+tg1536
467
tokens/s
32.7
tokens/s
59
ms